Earth2Studio.jl

Overview

Earth2Studio.jl wraps NVIDIA’s earth2studio Python package for AI weather and climate forecasting. It holds earth2studio and its submodules as PythonCall.jl Py objects, so the full Python API is callable from Julia. Nothing is exported: upstream names such as run and io would clash with Base.

Quickstart

On first using Earth2Studio, CondaPkg.jl installs Python and earth2studio[data]. Model extras are opt-in through a CondaPkg.toml in your own project:

[pip.deps.earth2studio]
extras = ["fcn", "pangu"]

See Basics for the Julia-to-Python conversion rules, Forecasting for running a model, and Julia packages for moving results into DimensionalData and Makie. The example below fetches a field from ARCO ERA5 and plots it with CairoMakie:

using Earth2Studio
using Dates
using CairoMakie
using PythonCall: pyconvert

ds = Earth2Studio.data.ARCO(verbose=false)
da = ds(DateTime(2023, 6, 15), "t2m")   # xarray.DataArray with dims (time, variable, lat, lon)

lon = pyconvert(Vector, da.lon.values)
lat = pyconvert(Vector, da.lat.values)
t2m = pyconvert(Array, da.values)[1, 1, :, :] .- 273.15   # lat × lon, °C

fig = Figure(size=(900, 450))
ax = Axis(fig[1, 1]; title="ERA5 2 m temperature, 2023-06-15 00:00 UTC", xlabel="Longitude", ylabel="Latitude")
hm = heatmap!(ax, lon, lat, permutedims(t2m))
Colorbar(fig[1, 2], hm; label="°C")
fig