CMIP6 quickstart — a bbox/time tas subset from the raw archive¶

earthlens.cmip6 exposes the full raw CMIP6 archive — the whole model × scenario × variable × member matrix — as analysis-ready Zarr on the open Pangeo gs://cmip6 bucket. Unlike the gee backend's single pre-downscaled NASA/GDDP-CMIP6 product, or the CHC-CMIP6 precipitation deltas, this reaches every ESM on its native grid.

This is a live, anonymous pull — no credentials. We resolve one store (model CanESM5, scenario ssp585, variable tas, monthly table Amon) and have pyramids write a small Europe / half-year NetCDF subset, then plot it.

Setup¶

EarthLens is the unified entry point; earthlens.cmip6.Catalog is the config + curated-vocabulary catalog. tempfile / pathlib give us a scratch directory for the written NetCDF.

In [1]:

import tempfile
from pathlib import Path

from earthlens import EarthLens
from earthlens.cmip6 import Catalog

import tempfile from pathlib import Path from earthlens import EarthLens from earthlens.cmip6 import Catalog

Explore the catalog (no network)¶

The bundled catalog carries the consolidated-stores CSV URL plus a curated vocabulary of common variables, experiments, and tables. Resolution runs against the full CSV, so uncurated facets still download — the curated rows just enrich metadata and error messages.

In [2]:

cat = Catalog()
print('variable :', cat.get_dataset('tas').long_name, '(', cat.get_dataset('tas').units, ')')
print('scenario :', 'ssp585 ->', cat.get_experiment('ssp585').description)
print('table    :', 'Amon ->', cat.get_table('Amon').description)
print('CSV index:', cat.csv_url)

cat = Catalog() print('variable :', cat.get_dataset('tas').long_name, '(', cat.get_dataset('tas').units, ')') print('scenario :', 'ssp585 ->', cat.get_experiment('ssp585').description) print('table :', 'Amon ->', cat.get_table('Amon').description) print('CSV index:', cat.csv_url)

variable : Near-surface (2 m) air temperature ( K )
scenario : ssp585 -> SSP5-8.5 — very high emissions, fossil-fuelled growth.
table    : Amon -> Monthly-mean atmospheric fields.
CSV index: https://storage.googleapis.com/cmip6/pangeo-cmip6.csv

Build and run the request¶

The request is a CMIP6 facet tuple plus a date window and a bounding box. We crop the native grid to Europe and the window to the first half of 2050, so pyramids fetches only those cells. download() returns the list[Path] of written NetCDF subsets — one per resolved store.

In [3]:

out_dir = Path(tempfile.mkdtemp(prefix='cmip6_'))
paths = EarthLens(
    'cmip6',
    source_id='CanESM5',
    experiment_id='ssp585',
    variable_id='tas',
    table_id='Amon',
    member_id='r1i1p1f1',
    start='2050-01-01',
    end='2050-06-30',
    lat_lim=[35.0, 60.0],
    lon_lim=[-10.0, 30.0],
    path=str(out_dir),
).download(progress_bar=False)
paths

out_dir = Path(tempfile.mkdtemp(prefix='cmip6_')) paths = EarthLens( 'cmip6', source_id='CanESM5', experiment_id='ssp585', variable_id='tas', table_id='Amon', member_id='r1i1p1f1', start='2050-01-01', end='2050-06-30', lat_lim=[35.0, 60.0], lon_lim=[-10.0, 30.0], path=str(out_dir), ).download(progress_bar=False) paths

2026-07-05 13:43:47 | INFO | pyramids.base.config | Logging is configured.

Out[3]:

[WindowsPath('C:/Users/main/AppData/Local/Temp/cmip6_vdj4q1mf/CanESM5_ssp585_tas_Amon_r1i1p1f1_gn_v20190429_20500101_20500630.nc')]

Inspect the written NetCDF¶

The read is entirely pyramids' (GDAL /vsigs/) — earthlens never imports xarray / zarr / gcsfs. The output carries one band per monthly timestep in the window (six bands for January–June 2050).

In [4]:

from pyramids.netcdf import NetCDF

nc = NetCDF.read_file(str(paths[0]))
print('bands (monthly steps):', len(nc.variable_names))
print(nc.variable_names)

from pyramids.netcdf import NetCDF nc = NetCDF.read_file(str(paths[0])) print('bands (monthly steps):', len(nc.variable_names)) print(nc.variable_names)

bands (monthly steps): 6
['Band1', 'Band2', 'Band3', 'Band4', 'Band5', 'Band6']

Plot the first month with pyramids¶

Each band is a georeferenced grid on CanESM5's native resolution. We pull the first month and plot it with pyramids (via cleopatra).

In [5]:

tas_jan = NetCDF.read_file(str(paths[0])).get_variable('Band1')
tas_jan.plot()

tas_jan = NetCDF.read_file(str(paths[0])).get_variable('Band1') tas_jan.plot()

Out[5]:

<cleopatra.array_glyph.ArrayGlyph at 0x2a3a0639be0>

Attribution¶

CMIP6 output is provided for research under each source model's terms (most are CC BY 4.0). Always cite the source GCM; the backend surfaces the per-model note.

In [6]:

backend = EarthLens(
    'cmip6', source_id='CanESM5', experiment_id='ssp585', variable_id='tas',
    table_id='Amon', start='2050-01-01', end='2050-06-30',
    lat_lim=[35.0, 60.0], lon_lim=[-10.0, 30.0], path=str(out_dir),
).datasource
print(backend.terms_note())

backend = EarthLens( 'cmip6', source_id='CanESM5', experiment_id='ssp585', variable_id='tas', table_id='Amon', start='2050-01-01', end='2050-06-30', lat_lim=[35.0, 60.0], lon_lim=[-10.0, 30.0], path=str(out_dir), ).datasource print(backend.terms_note())

CC BY 4.0 — cite CanESM5 (Swart et al., 2019).