Creating a Local Data Source#

Create and save an offline dataset to use in an inference pipeline.

This example demonstrates how to:

  • Build a small offline dataset by fetching data and writing to a Zarr store

  • Load the local store as a data source for an inference pipeline with the Microsoft Aurora model

  • Run the deterministic workflow and plot results

# /// script
# dependencies = [
#   "earth2studio @ git+https://github.com/NVIDIA/earth2studio.git",
#   "aurora",
#   "cartopy",
# ]
# ///

Set Up#

For this example, the following are needed:

import os

os.makedirs("outputs", exist_ok=True)
from dotenv import load_dotenv

load_dotenv()

from earth2studio.data import WB2ERA5, fetch_data
from earth2studio.models.px import Aurora

# Load the default model package which downloads the checkpoint from GCP
package = Aurora.load_default_package()
model = Aurora.load_model(package)

# Create the data source, cache is false
wb2 = WB2ERA5(cache=False, verbose=False)

Creating a Local Zarr Store from a Datasource#

Start with creating a local dataset from the WeatherBench2 data store. Since data sources return in-memory data arrays, there are a variety of ways this could be done. The following is a simple method using Earth2Studio IO objects to pack the requested data into a single Zarr store.

For this example, let’s download some data for a Microsoft aurora forecast.

from collections import OrderedDict

import numpy as np

from earth2studio.io import ZarrBackend
from earth2studio.utils.coords import split_coords

times = np.array(
    [np.datetime64("2022-01-01T00:00:00"), np.datetime64("2022-01-01T06:00:00")]
)
variables = model.input_coords()["variable"]
zarr_path = "./outputs/19_wb2_dataset.zarr"
# Create Zarr store to pack data into
zb = ZarrBackend(file_name=zarr_path, backend_kwargs={"overwrite": True})
full_coords = OrderedDict(
    [
        ("time", np.atleast_1d(times)),
        ("lead_time", np.array([np.timedelta64(0, "h")])),
        ("lat", np.linspace(90, -90, 721)),
        ("lon", np.linspace(0, 359.75, 1440)),
    ]
)
zb.add_array(full_coords, array_name=list(variables))

# Loop over timestamps, fetch data and write slices into the pre-created arrays
for t in np.atleast_1d(times):
    x, coords = fetch_data(
        wb2,
        time=np.array([t]),
        variable=variables,
        lead_time=np.array([np.timedelta64(0, "h")]),
        device="cpu",
    )
    xs, reduced_coords, var_names = split_coords(x, coords, dim="variable")
    zb.write(xs, reduced_coords, array_name=list(var_names))
2025-12-15 01:40:47.530 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t300 at 2022-01-01T00:00:00
2025-12-15 01:40:47.531 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q850 at 2022-01-01T00:00:00
2025-12-15 01:40:47.531 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t2m at 2022-01-01T00:00:00
2025-12-15 01:40:47.532 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t700 at 2022-01-01T00:00:00
2025-12-15 01:40:47.532 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q700 at 2022-01-01T00:00:00
2025-12-15 01:40:47.533 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t1000 at 2022-01-01T00:00:00
2025-12-15 01:40:47.533 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u700 at 2022-01-01T00:00:00
2025-12-15 01:40:47.533 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q200 at 2022-01-01T00:00:00
2025-12-15 01:40:47.534 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z700 at 2022-01-01T00:00:00
2025-12-15 01:40:47.534 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u200 at 2022-01-01T00:00:00
2025-12-15 01:40:47.535 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u50 at 2022-01-01T00:00:00
2025-12-15 01:40:47.535 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z500 at 2022-01-01T00:00:00
2025-12-15 01:40:47.536 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t925 at 2022-01-01T00:00:00
2025-12-15 01:40:47.536 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z300 at 2022-01-01T00:00:00
2025-12-15 01:40:47.537 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q500 at 2022-01-01T00:00:00
2025-12-15 01:40:47.537 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z850 at 2022-01-01T00:00:00
2025-12-15 01:40:47.537 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q400 at 2022-01-01T00:00:00
2025-12-15 01:40:47.538 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u925 at 2022-01-01T00:00:00
2025-12-15 01:40:47.538 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v300 at 2022-01-01T00:00:00
2025-12-15 01:40:47.539 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v200 at 2022-01-01T00:00:00
2025-12-15 01:40:47.539 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q150 at 2022-01-01T00:00:00
2025-12-15 01:40:47.539 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t400 at 2022-01-01T00:00:00
2025-12-15 01:40:47.540 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z1000 at 2022-01-01T00:00:00
2025-12-15 01:40:47.540 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: msl at 2022-01-01T00:00:00
2025-12-15 01:40:47.541 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z925 at 2022-01-01T00:00:00
2025-12-15 01:40:47.541 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q100 at 2022-01-01T00:00:00
2025-12-15 01:40:47.542 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u400 at 2022-01-01T00:00:00
2025-12-15 01:40:47.542 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t600 at 2022-01-01T00:00:00
2025-12-15 01:40:47.542 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u300 at 2022-01-01T00:00:00
2025-12-15 01:40:47.543 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z50 at 2022-01-01T00:00:00
2025-12-15 01:40:47.543 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v600 at 2022-01-01T00:00:00
2025-12-15 01:40:47.544 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q1000 at 2022-01-01T00:00:00
2025-12-15 01:40:47.544 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z250 at 2022-01-01T00:00:00
2025-12-15 01:40:47.545 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v50 at 2022-01-01T00:00:00
2025-12-15 01:40:47.545 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v500 at 2022-01-01T00:00:00
2025-12-15 01:40:47.545 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t500 at 2022-01-01T00:00:00
2025-12-15 01:40:47.546 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v400 at 2022-01-01T00:00:00
2025-12-15 01:40:47.546 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u10m at 2022-01-01T00:00:00
2025-12-15 01:40:47.547 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t200 at 2022-01-01T00:00:00
2025-12-15 01:40:47.547 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t250 at 2022-01-01T00:00:00
2025-12-15 01:40:47.548 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z100 at 2022-01-01T00:00:00
2025-12-15 01:40:47.548 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u1000 at 2022-01-01T00:00:00
2025-12-15 01:40:47.548 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q300 at 2022-01-01T00:00:00
2025-12-15 01:40:47.549 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v150 at 2022-01-01T00:00:00
2025-12-15 01:40:47.549 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t50 at 2022-01-01T00:00:00
2025-12-15 01:40:47.550 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z400 at 2022-01-01T00:00:00
2025-12-15 01:40:47.550 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z600 at 2022-01-01T00:00:00
2025-12-15 01:40:47.550 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q250 at 2022-01-01T00:00:00
2025-12-15 01:40:47.551 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u600 at 2022-01-01T00:00:00
2025-12-15 01:40:47.551 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u100 at 2022-01-01T00:00:00
2025-12-15 01:40:47.552 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v100 at 2022-01-01T00:00:00
2025-12-15 01:40:47.552 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q925 at 2022-01-01T00:00:00
2025-12-15 01:40:47.553 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t100 at 2022-01-01T00:00:00
2025-12-15 01:40:47.553 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u500 at 2022-01-01T00:00:00
2025-12-15 01:40:47.553 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q600 at 2022-01-01T00:00:00
2025-12-15 01:40:47.554 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q50 at 2022-01-01T00:00:00
2025-12-15 01:40:47.554 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v925 at 2022-01-01T00:00:00
2025-12-15 01:40:47.555 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v10m at 2022-01-01T00:00:00
2025-12-15 01:40:47.555 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u150 at 2022-01-01T00:00:00
2025-12-15 01:40:47.556 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v850 at 2022-01-01T00:00:00
2025-12-15 01:40:47.556 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v250 at 2022-01-01T00:00:00
2025-12-15 01:40:47.556 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t850 at 2022-01-01T00:00:00
2025-12-15 01:40:47.557 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v700 at 2022-01-01T00:00:00
2025-12-15 01:40:47.557 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u250 at 2022-01-01T00:00:00
2025-12-15 01:40:47.558 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t150 at 2022-01-01T00:00:00
2025-12-15 01:40:47.558 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v1000 at 2022-01-01T00:00:00
2025-12-15 01:40:47.558 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z150 at 2022-01-01T00:00:00
2025-12-15 01:40:47.559 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z200 at 2022-01-01T00:00:00
2025-12-15 01:40:47.559 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u850 at 2022-01-01T00:00:00
2025-12-15 01:40:56.813 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q1000 at 2022-01-01T06:00:00
2025-12-15 01:40:56.814 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v1000 at 2022-01-01T06:00:00
2025-12-15 01:40:56.814 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v400 at 2022-01-01T06:00:00
2025-12-15 01:40:56.815 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q500 at 2022-01-01T06:00:00
2025-12-15 01:40:56.815 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q250 at 2022-01-01T06:00:00
2025-12-15 01:40:56.816 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t500 at 2022-01-01T06:00:00
2025-12-15 01:40:56.816 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q50 at 2022-01-01T06:00:00
2025-12-15 01:40:56.817 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t2m at 2022-01-01T06:00:00
2025-12-15 01:40:56.817 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z500 at 2022-01-01T06:00:00
2025-12-15 01:40:56.818 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z400 at 2022-01-01T06:00:00
2025-12-15 01:40:56.818 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t150 at 2022-01-01T06:00:00
2025-12-15 01:40:56.818 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q300 at 2022-01-01T06:00:00
2025-12-15 01:40:56.819 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v500 at 2022-01-01T06:00:00
2025-12-15 01:40:56.819 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z150 at 2022-01-01T06:00:00
2025-12-15 01:40:56.820 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: msl at 2022-01-01T06:00:00
2025-12-15 01:40:56.820 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u1000 at 2022-01-01T06:00:00
2025-12-15 01:40:56.821 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u10m at 2022-01-01T06:00:00
2025-12-15 01:40:56.821 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q100 at 2022-01-01T06:00:00
2025-12-15 01:40:56.821 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q850 at 2022-01-01T06:00:00
2025-12-15 01:40:56.822 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v250 at 2022-01-01T06:00:00
2025-12-15 01:40:56.822 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u150 at 2022-01-01T06:00:00
2025-12-15 01:40:56.823 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q150 at 2022-01-01T06:00:00
2025-12-15 01:40:56.823 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z250 at 2022-01-01T06:00:00
2025-12-15 01:40:56.823 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z300 at 2022-01-01T06:00:00
2025-12-15 01:40:56.824 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z700 at 2022-01-01T06:00:00
2025-12-15 01:40:56.824 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z600 at 2022-01-01T06:00:00
2025-12-15 01:40:56.825 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v50 at 2022-01-01T06:00:00
2025-12-15 01:40:56.825 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u500 at 2022-01-01T06:00:00
2025-12-15 01:40:56.826 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z1000 at 2022-01-01T06:00:00
2025-12-15 01:40:56.826 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u200 at 2022-01-01T06:00:00
2025-12-15 01:40:56.826 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v850 at 2022-01-01T06:00:00
2025-12-15 01:40:56.827 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v10m at 2022-01-01T06:00:00
2025-12-15 01:40:56.827 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u300 at 2022-01-01T06:00:00
2025-12-15 01:40:56.828 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v150 at 2022-01-01T06:00:00
2025-12-15 01:40:56.828 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q700 at 2022-01-01T06:00:00
2025-12-15 01:40:56.828 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u400 at 2022-01-01T06:00:00
2025-12-15 01:40:56.829 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v300 at 2022-01-01T06:00:00
2025-12-15 01:40:56.829 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u100 at 2022-01-01T06:00:00
2025-12-15 01:40:56.830 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t925 at 2022-01-01T06:00:00
2025-12-15 01:40:56.830 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q925 at 2022-01-01T06:00:00
2025-12-15 01:40:56.830 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z850 at 2022-01-01T06:00:00
2025-12-15 01:40:56.831 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t700 at 2022-01-01T06:00:00
2025-12-15 01:40:56.831 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t100 at 2022-01-01T06:00:00
2025-12-15 01:40:56.832 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z100 at 2022-01-01T06:00:00
2025-12-15 01:40:56.832 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t200 at 2022-01-01T06:00:00
2025-12-15 01:40:56.833 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z200 at 2022-01-01T06:00:00
2025-12-15 01:40:56.833 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t1000 at 2022-01-01T06:00:00
2025-12-15 01:40:56.833 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v700 at 2022-01-01T06:00:00
2025-12-15 01:40:56.834 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u250 at 2022-01-01T06:00:00
2025-12-15 01:40:56.834 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v200 at 2022-01-01T06:00:00
2025-12-15 01:40:56.835 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z50 at 2022-01-01T06:00:00
2025-12-15 01:40:56.835 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t400 at 2022-01-01T06:00:00
2025-12-15 01:40:56.836 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q200 at 2022-01-01T06:00:00
2025-12-15 01:40:56.837 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u50 at 2022-01-01T06:00:00
2025-12-15 01:40:56.837 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u850 at 2022-01-01T06:00:00
2025-12-15 01:40:56.838 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t250 at 2022-01-01T06:00:00
2025-12-15 01:40:56.838 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t50 at 2022-01-01T06:00:00
2025-12-15 01:40:56.838 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u700 at 2022-01-01T06:00:00
2025-12-15 01:40:56.839 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v925 at 2022-01-01T06:00:00
2025-12-15 01:40:56.839 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v600 at 2022-01-01T06:00:00
2025-12-15 01:40:56.840 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q400 at 2022-01-01T06:00:00
2025-12-15 01:40:56.840 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u600 at 2022-01-01T06:00:00
2025-12-15 01:40:56.840 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t300 at 2022-01-01T06:00:00
2025-12-15 01:40:56.841 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t850 at 2022-01-01T06:00:00
2025-12-15 01:40:56.841 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: u925 at 2022-01-01T06:00:00
2025-12-15 01:40:56.842 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: z925 at 2022-01-01T06:00:00
2025-12-15 01:40:56.842 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: v100 at 2022-01-01T06:00:00
2025-12-15 01:40:56.843 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: q600 at 2022-01-01T06:00:00
2025-12-15 01:40:56.843 | DEBUG    | earth2studio.data.wb2:fetch_array:241 - Fetching WB2 zarr array for variable: t600 at 2022-01-01T06:00:00

Note that the Zarr store we just created can be used for more than just Earth2Studio inference pipelines. Open it with zarr or xarray to explore/process what you just downloaded.

import zarr

zg = zarr.group(store=zarr.storage.LocalStore(zarr_path))
print(zg.tree())
/
├── lat (721,) float64
├── lead_time (1,) timedelta64
├── lon (1440,) float64
├── msl (2, 1, 721, 1440) float32
├── q100 (2, 1, 721, 1440) float32
├── q1000 (2, 1, 721, 1440) float32
├── q150 (2, 1, 721, 1440) float32
├── q200 (2, 1, 721, 1440) float32
├── q250 (2, 1, 721, 1440) float32
├── q300 (2, 1, 721, 1440) float32
├── q400 (2, 1, 721, 1440) float32
├── q50 (2, 1, 721, 1440) float32
├── q500 (2, 1, 721, 1440) float32
├── q600 (2, 1, 721, 1440) float32
├── q700 (2, 1, 721, 1440) float32
├── q850 (2, 1, 721, 1440) float32
├── q925 (2, 1, 721, 1440) float32
├── t100 (2, 1, 721, 1440) float32
├── t1000 (2, 1, 721, 1440) float32
├── t150 (2, 1, 721, 1440) float32
├── t200 (2, 1, 721, 1440) float32
├── t250 (2, 1, 721, 1440) float32
├── t2m (2, 1, 721, 1440) float32
├── t300 (2, 1, 721, 1440) float32
├── t400 (2, 1, 721, 1440) float32
├── t50 (2, 1, 721, 1440) float32
├── t500 (2, 1, 721, 1440) float32
├── t600 (2, 1, 721, 1440) float32
├── t700 (2, 1, 721, 1440) float32
├── t850 (2, 1, 721, 1440) float32
├── t925 (2, 1, 721, 1440) float32
├── time (2,) datetime64
├── u100 (2, 1, 721, 1440) float32
├── u1000 (2, 1, 721, 1440) float32
├── u10m (2, 1, 721, 1440) float32
├── u150 (2, 1, 721, 1440) float32
├── u200 (2, 1, 721, 1440) float32
├── u250 (2, 1, 721, 1440) float32
├── u300 (2, 1, 721, 1440) float32
├── u400 (2, 1, 721, 1440) float32
├── u50 (2, 1, 721, 1440) float32
├── u500 (2, 1, 721, 1440) float32
├── u600 (2, 1, 721, 1440) float32
├── u700 (2, 1, 721, 1440) float32
├── u850 (2, 1, 721, 1440) float32
├── u925 (2, 1, 721, 1440) float32
├── v100 (2, 1, 721, 1440) float32
├── v1000 (2, 1, 721, 1440) float32
├── v10m (2, 1, 721, 1440) float32
├── v150 (2, 1, 721, 1440) float32
├── v200 (2, 1, 721, 1440) float32
├── v250 (2, 1, 721, 1440) float32
├── v300 (2, 1, 721, 1440) float32
├── v400 (2, 1, 721, 1440) float32
├── v50 (2, 1, 721, 1440) float32
├── v500 (2, 1, 721, 1440) float32
├── v600 (2, 1, 721, 1440) float32
├── v700 (2, 1, 721, 1440) float32
├── v850 (2, 1, 721, 1440) float32
├── v925 (2, 1, 721, 1440) float32
├── z100 (2, 1, 721, 1440) float32
├── z1000 (2, 1, 721, 1440) float32
├── z150 (2, 1, 721, 1440) float32
├── z200 (2, 1, 721, 1440) float32
├── z250 (2, 1, 721, 1440) float32
├── z300 (2, 1, 721, 1440) float32
├── z400 (2, 1, 721, 1440) float32
├── z50 (2, 1, 721, 1440) float32
├── z500 (2, 1, 721, 1440) float32
├── z600 (2, 1, 721, 1440) float32
├── z700 (2, 1, 721, 1440) float32
├── z850 (2, 1, 721, 1440) float32
└── z925 (2, 1, 721, 1440) float32

Execute the Workflow#

To use the saved dataset as a data source, we could create our own class that implements the interface required by earth2studio.data.base.DataSource, which needs just a __call__(time, variable) method.

However, since we used an IO backend from Earth2Studio we can use the earth2studio.data.xr.InferenceOutputSource which is a convenience class that supports the output of inference pipelines.

import earth2studio.run as run
from earth2studio.data import InferenceOutputSource

offline_source = InferenceOutputSource(zarr_path)
out_zarr_path = "./outputs/19_pangu_output.zarr"
io = ZarrBackend(file_name=out_zarr_path, backend_kwargs={"overwrite": True})
io = run.deterministic(
    times[-1:],
    4,
    model,
    offline_source,
    io,
    output_coords=OrderedDict({"variable": np.array(["msl"])}),
)
2025-12-15 01:41:12.091 | INFO     | earth2studio.run:deterministic:75 - Running simple workflow!
2025-12-15 01:41:12.092 | INFO     | earth2studio.run:deterministic:82 - Inference device: cuda
2025-12-15 01:41:13.227 | SUCCESS  | earth2studio.run:deterministic:106 - Fetched data from InferenceOutputSource
2025-12-15 01:41:13.258 | INFO     | earth2studio.run:deterministic:136 - Inference starting!


Running inference:   0%|          | 0/5 [00:00<?, ?it/s]

Running inference:  40%|████      | 2/5 [00:04<00:07,  2.50s/it]

Running inference:  60%|██████    | 3/5 [00:09<00:07,  3.51s/it]

Running inference:  80%|████████  | 4/5 [00:14<00:04,  4.06s/it]

Running inference: 100%|██████████| 5/5 [00:19<00:00,  4.37s/it]
Running inference: 100%|██████████| 5/5 [00:19<00:00,  3.97s/it]
2025-12-15 01:41:33.105 | SUCCESS  | earth2studio.run:deterministic:146 - Inference complete

Post Processing#

The last step is to post-process our results.

import cartopy.crs as ccrs
import matplotlib.pyplot as plt

plt.close("all")
projection = ccrs.Robinson()
fig, axes = plt.subplots(
    2,
    2,
    subplot_kw={"projection": projection},
    figsize=(12, 7),
    constrained_layout=True,
)
axes = axes.ravel()

lon = io["lon"][:]
lat = io["lat"][:]
lead_steps = [1, 2, 3, 4]  # 6h, 12h, 18h, 24h
for ax, step in zip(axes, lead_steps):
    im = ax.pcolormesh(
        lon,
        lat,
        io["msl"][0, step],
        transform=ccrs.PlateCarree(),
        cmap="PiYG",
    )
    ax.set_title(f"msl - Lead time: {6*step}h")
    ax.coastlines()
    ax.gridlines(draw_labels=False)

fig.colorbar(
    im, ax=axes, orientation="horizontal", fraction=0.05, pad=0.07, label="msl"
)
plt.savefig("outputs/19_msl_1day.png", dpi=150)
msl - Lead time: 6h, msl - Lead time: 12h, msl - Lead time: 18h, msl - Lead time: 24h

Total running time of the script: (1 minutes 59.607 seconds)

Gallery generated by Sphinx-Gallery