Installation

Installation#

Get radar-datatree running locally in two minutes.

Prerequisites#

  • Python ≥ 3.12 — required by xarray ≥ 2026.4 and the rest of the stack.

  • Network access to s3://nexrad-arco on AWS us-east-1 (anonymous reads — no AWS credentials needed).

Install#

We strongly recommend uv — a fast Python package manager — but conda works too.

git clone https://github.com/AtmoScale/radar-datatree.git
cd radar-datatree
uv sync
uv run jupyter lab notebooks/

git clone https://github.com/AtmoScale/radar-datatree.git
cd radar-datatree
conda env create -f environment.yml
conda activate radar-datatree
jupyter lab notebooks/

No multi-gigabyte downloads required — data streams directly from the cloud.

Verify your install#

Run the Quickstart snippet in a Python session (or a fresh notebook cell). If it prints a list of VCP-* nodes, your install is healthy — every dependency loaded, anonymous S3 reads work, and you’re already streaming from the public archive. Then jump straight into Notebook 1 for the full polarimetric walkthrough.

What got installed#

Core radar-datatree stack

Package

Version

Why it matters

xarray

≥ 2026.4

Hierarchical DataTree data model.

xradar

≥ 0.12

Radar-specific I/O on top of xarray (CfRadial, NEXRAD Level II readers, QC).

zarr

≥ 3.1.2

Cloud-optimized chunked storage; required for the v3 spec used in nexrad-arco.

icechunk

≥ 2.0.4

ACID-compliant transactional layer over Zarr — opens the archive as a versioned repo.

rustytree-xarray

≥ 0.2

Rust-backed DataTree backend (engine="rustytree"); ~10× faster than engine="zarr" on icechunk repos served from object storage.

s3fs

≥ 2025.5.1

Anonymous S3 reads against the AWS Open Data bucket.

Full dependency tree (including visualization extras for individual tutorials) lives in pyproject.toml.

Contents