am-python

Python bindings for Scott Paine’s am atmospheric model, via Rust (PyO3).

Install

Pre-built wheels are available on PyPI for Linux:

uv pip install am-python

From source

Requires the am source code (tested on v14.0), a C compiler, and a Rust toolchain.

curl -fsSL "https://zenodo.org/records/13748403/files/am-14.0.tgz?download=1" | tar -xz
export AM_SRC_DIR=$PWD/am-14.0/src
uv pip install .

Usage

Single model

import am

m = am.Model("SPole_JJA_75.amc", [0, "GHz", 350, "GHz", 0.01, "GHz", 35, "deg", 1.0])
m.compute()

# numpy array, GHz
m.frequency
 # dict of computed outputs
m.outputs
# Rayleigh-Jeans brightness temperature, K
m.outputs["tb_rj"]

Parameter sweep

import xarray as xr
import am

params = xr.Dataset(coords={
    "elevation": [30, 45, 60, 75],   # deg
    "pwv":       [0.5, 1.0, 2.0],    # mm
})

ds = am.ModelGrid(
    "SPole_JJA_75.amc",
    params,
    args_fn=lambda elevation, pwv: [
        0, "GHz", 350, "GHz", 0.5, "GHz", elevation, "deg", pwv
    ],
).compute()

# xr.Dataset with dims (elevation, pwv, frequency)
ds["tb_rj"].sel(elevation=45, pwv=1.0)

For irregular grids (e.g. a set of specific elevation/PWV pairs), use variables on a shared dimension instead:

import numpy as np

params = xr.Dataset({
    "elevation": ("obs", elev_array),
    "pwv":       ("obs", pwv_array),
})

Development

This project uses just to orchestrate common tasks and pre-commit for local checks.

# install deps
just sync

# run tests
just test
just test py
just test rs

# run formatting/lint checks
just fmt
just fmt-check
just lint
just typecheck

# build docs
just docs
just docs py
just docs rs

# run pre-commit hooks on all files
just precommit
just prepush

Internal Rust crate documentation is available here.