How to Download and Analyse ERA5 in Python with xarray (New CDS, 2026)

ERA5 is the dataset most climate analysts reach for first: consistent, global and free. Getting it into Python used to be well documented, but the Climate Data Store changed in 2024, and many tutorials still show the old key format and the old file layout. This walkthrough uses the current system, from setting up the key to a chart of annual temperature anomalies for any area.

Step 1: account, licence and key

Register at the Climate Data Store. Open the ERA5 dataset page and accept its licence once, otherwise every request fails. Then copy your personal access token from your profile and save it in a file called .cdsapirc in your home folder:

url: https://cds.climate.copernicus.eu/api
key: YOUR-PERSONAL-ACCESS-TOKEN

Old guides show a key in the form “UID:API-key”. That format no longer works.

Step 2: download monthly 2 m temperature

# pip install "cdsapi>=0.7.2" xarray netCDF4 matplotlib
import cdsapi

client = cdsapi.Client()
client.retrieve(
    "reanalysis-era5-single-levels-monthly-means",
    {
        "product_type": ["monthly_averaged_reanalysis"],
        "variable": ["2m_temperature"],
        "year": [str(y) for y in range(1991, 2026)],
        "month": [f"{m:02d}" for m in range(1, 13)],
        "time": ["00:00"],
        "area": [52, -1, 51, 1],  # north, west, south, east (here: around London)
        "data_format": "netcdf",
        "download_format": "unarchived",
    },
    "era5_t2m_monthly.nc",
)

Requests join a queue, so a small one like this can still take a few minutes. Change the area to your own; keep it small while you learn.

Step 3: open it, and handle the new layout

import zipfile
import numpy as np
import xarray as xr

path = "era5_t2m_monthly.nc"
if zipfile.is_zipfile(path):            # some requests still arrive zipped
    with zipfile.ZipFile(path) as z:
        z.extractall("era5_files")
        path = "era5_files/" + z.namelist()[0]

ds = xr.open_dataset(path)
print(ds)                               # check the dimension names first

if "valid_time" in ds.dims:             # new CDS files
    ds = ds.rename({"valid_time": "time"})

t2m = ds["t2m"] - 273.15                # kelvin to degrees Celsius

Always print the dataset once. If you request several variables, accumulated ones such as precipitation and instantaneous ones such as temperature can come back as separate files inside the zip; open each one.

Step 4: area average and anomalies

weights = np.cos(np.deg2rad(t2m.latitude))
series = t2m.weighted(weights).mean(["latitude", "longitude"])

annual = series.resample(time="YS").mean()
baseline = annual.sel(time=slice("1991", "2020")).mean()
anomaly = annual - baseline

ax = anomaly.plot.line(marker="o")
ax[0].axes.axhline(0, color="grey", linewidth=0.8)
ax[0].axes.set_ylabel("°C vs 1991–2020")
ax[0].axes.set_title("Annual mean 2 m temperature anomaly")

The cosine weights matter for larger areas, because grid cells shrink towards the poles. The 1991–2020 baseline matches the current World Meteorological Organization climate normal, so your numbers are comparable with published anomalies. Leave out the current year if it is not yet complete.

Where to go next

The same pattern works for rainfall (total_precipitation, in metres), wind and soil moisture. For hourly data, use reanalysis-era5-single-levels and request only the hours and days you need; for land-only detail at about 9 km, try ERA5-Land. For climate projections rather than the past, the CDS also hosts CMIP6 data, which xarray opens the same way.