OPEN DATA · PYTHON SDK + CLI

Public science.
Reproducible inputs.

usdata is for anyone who puts federal data, from NOAA weather to EPA air quality, into a paper, a model, or a report and has to show where every number came from. One command finds the dataset, one fetches the exact source file, one opens it, and a small manifest lets anyone repeat the download and prove the bytes match.

Start with one dataset Browse datasets

YOUR TERMINAL

python -m pip install "usdata[pandas]"
usdata search precipitation --location Oklahoma
usdata fetch noaa:ghcn-daily -p stations=USW00013967 --start 2024-05-06 --end 2024-05-07 --vars PRCP,TMAX
usdata pull dataset.yaml && usdata verify dataset.yaml

ONE RESULT, END TO END

How high was Hurricane Helene's storm surge at Cedar Key?

Three fetches: six-minute water levels and tide predictions for one NOAA gauge, and the hurricane's best track. Subtracting the predictions from the observations isolates the surge.

Peak residual: 3.15 m above the predicted tide
When: 04:54 UTC, 27 September 2024, 1.7 h after landfall
Highest water: 3.99 m above mean lower low water

Every input is pinned. The notebook ends by checking the cached bytes against the lockfile, restoring from it, and citing each source.

noaa:coops-water-levels
Bytes: 28075; cache hit: False
Checksum: sha256:431fade50c78a7f1ea1f4bf27f40a54fca8809d723d26eb5b4bd7c98623057b2

$ usdata verify dataset.yaml
all assets match dataset.lock.json

Open the storm-surge study

Two panels: observed and predicted water level at Cedar Key over three days, and the residual between them peaking near 3.15 metres early on 27 September 2024. — Observed water level, predicted tide, and their difference at NOAA station 8727520, from the study's saved output.

Find your source

A curated catalog of datasets you can fetch today, each with its formats, selection rules, and required inputs stated up front.

Browse datasets

Fetch and open

Download the source files unchanged, with a provenance sidecar beside each one. Optional readers open CSV, radar, NetCDF, and GRIB2 into pandas or xarray.

Follow the guide

Keep the evidence

A manifest names the query; a lockfile pins the bytes. Restore on another machine, and find out when an agency quietly revises a file.

Use a manifest

STUDIES

Questions with saved answers

Each study is a notebook that ran against the live services, with its outputs, plots, and exact inputs kept. Every dataset also has its own walkthrough on its page.