GHCN-Daily: global daily weather observations
registry.opendata.aws/noaa-ghcn
The global daily weather observation archive NOAA publishes as the Global Historical Climatology Network – Daily, kept in the source’s own long form — one row per station, day and element: maximum and minimum temperature, precipitation, snow, wind, pressure, humidity, sunshine, soil temperature and weather types. About 3.2 billion measurements from 132,501 stations worldwide, some reaching back to 1750, plus the 782,538-row inventory saying which years and elements each station covers. REFERENCED LIVE, NOT COPIED: five of the six tables — observations, stations, inventory, countries, states — are published as views over NOAA’s own public AWS Open Data bucket, so no row of them is stored in this directory’s lake and every query reads NOAA directly. That means the data is always exactly what NOAA publishes right now, which matters more here than for most sources: NOAA reprocesses its ENTIRE period of record every day, so a stored copy would freeze yesterday’s revision of an archive that runs from 1763 to today. It also means there is no history of what a value used to say before NOAA revised it, no row counts or column statistics of our own, and a query needs network access to s3://noaa-ghcn-pds (public, no credentials). The sixth table, elements, IS stored here as ordinary parquet — all 213 documented measurement types rather than a curated handful — because it is transcribed from the source’s prose readme and has no upstream file to read live. Every observation row carries both the integer NOAA published and the value converted to its real unit, alongside the quality, measurement and source flags that say how far to trust it.
CC0-1.0Creative Commons Zero v1.0 Universal (US Government work, public domain)
Attribution, terms and provenanceWhat you must carry if you reuse this
- Attribution
- Data from the NOAA National Centers for Environmental Information, Global Historical Climatology Network – Daily (GHCN-Daily), Version 3. Menne, M.J., I. Durre, B. Korzeniewski, S. McNeill, K. Thomas, X. Yin, S. Anthony, R. Ray, R.S. Vose, B.E. Gleason and T.G. Houston, 2012. doi:10.7289/V5D21VHZ.
- Notes
- GHCN-Daily is a work of the United States federal government, produced by NOAA’s National Centers for Environmental Information, and US federal works carry no copyright under 17 U.S.C. §105. NOAA’s data policy places its data in the public domain and asks for acknowledgement rather than requiring it; the AWS Open Data registry entry for s3://noaa-ghcn-pds states the same. Redistribution is therefore permitted, and CC0-1.0 is the closest SPDX identifier to "no rights reserved, attribution requested". IMPORTANT CAVEATS FOR A REUSER. (1) GHCN-Daily is an AGGREGATE: it merges data from many national and regional meteorological services — Environment Canada, the Australian Bureau of Meteorology, the European Climate Assessment and Dataset, the China Meteorological Administration, CONAGUA, the Fiji Met Service, several African services whose data was previously withheld pending permission, and about twenty more, each identified by the source_flag column. Those providers’ own terms may differ from NOAA’s for their contribution, and a reuser redistributing a subset should check the flag. NOAA publishes the merged product openly, and that is what is republished here. (2) Values that FAILED a quality check are published, not removed: a non-null quality_flag means the value failed the check that letter names. Presenting flagged values as measurements without saying so misrepresents the source. (3) Values are model-free observations, but TAVG from source flag "S" is derived from hourly synoptic reports and can differ significantly from a true daily mean, particularly for precipitation — the readme says so explicitly. (4) What is republished is a derived work: the source’s own columns, reshaped, with a scaled value and the element catalogue added. (5) For five of the six tables nothing is redistributed at all — they are published as views over NOAA’s own bucket, so what is offered is the reshaping, not the bytes, and a reader’s query fetches the data from NOAA under NOAA’s terms directly. Only `elements`, transcribed from the source readme, is actually stored and served here.
- Version
- snapshot 2026-10-22 · 01M5GS2HENV35C87PAWCVAV4ZY
- Machine-readable:
- llms.txtmanifest.json
Tables
Open a table for its columnsLoading the schema…
Queries
Copy a snippet, or run it in the playgroundQuery it yourselfConnection snippets for DuckDB — Parquet or the live lake
Generated from this exact manifest version, so the URLs are the ones actually published.
ATTACH IF NOT EXISTS 'ducklake:https://data-directory.fsn1.your-objectstorage.com/catalog/catalog.ducklake' AS "datadirectory" (READ_ONLY);
CREATE OR REPLACE SECRET "noaa_ghcn_pds" (TYPE s3, SCOPE 's3://noaa-ghcn-pds', PROVIDER 'config', REGION 'us-east-1');
ATTACH IF NOT EXISTS ':memory:' AS "ghcn_daily";
CREATE SCHEMA IF NOT EXISTS "ghcn_daily"."main";
CREATE OR REPLACE VIEW "ghcn_daily"."main"."elements" AS
SELECT *
FROM "datadirectory"."ghcn_daily"."elements";
SELECT *
FROM ghcn_daily.main.elements
LIMIT 10;