Source and reproduction record

Preview note: archived third-party HTML and extracted page text remain in the source repository. This preview provides the original source links, the frozen CSV, retrieval metadata and reproduction files.

Snapshot retrieved 15 October 2026 (UTC). Exact retrieval timestamps, URLs, HTTP metadata, sizes and SHA-256 hashes are in retrievals.json. HTML files are frozen reference evidence; adjacent .txt files are extracted page text for inspection. They are not chart inputs.

Data

• C3S September 2026 global monthly surface-air-temperature CSV, saved unchanged as global-monthly.csv.
• SHA-256: b31271d6984e7e4cc02bedffd268e68d7b5cda571fb521dcedf435c2eca5407e; 51,706 bytes; source header updated 3 October 2026; server Last-Modified 9 October 2026 13:29:22 GMT. Re-fetched and hash-checked 15 October 2026 at 20:38:16 UTC.
• Publisher: Copernicus Climate Change Service (C3S), implemented by ECMWF. The CSV describes globally averaged ERA5 surface air temperature, January 1940–September 2026, in degrees Celsius. Its original comment metadata is retained.
• Underlying product: ERA5 monthly averaged data on single levels from 1940 to present, DOI 10.24381/cds.f17050d7. Catalog snapshot era5-catalog.html identifies CC-BY licence; licence page is saved as cc-by.html, linking to CC BY 4.0.
• Attribution for charts and derived analysis: Data: C3S/ECMWF, ERA5 (September 2026 bulletin snapshot); calculations: DataHub. Contains modified Copernicus Climate Change Service information 2026. Modifications: monthly grouping, arithmetic averaging under two stated temporal weighting conventions, historical rankings, and conditional calculations. No source observations have been altered.
• Rights evidence concerns reuse of the ERA5 data and derived calculations. Reference-page HTML is retained as evidence; it is not asserted to share the dataset licence or proposed as a republished website.

Primary interpretive evidence

• September 2026 temperature bulletin: bulletin.html, text. Global temperature/year-to-date sections identify September as second warmest; January–September 2026 behind 2024 and ahead of 2023/2025; repeating the last three months of 2023 would leave 2026 close to 2024. It also gives C3S's El Niño context, which the calculation does not test.
• C3S news release: news.html, text. Page displays 12 October 2026 and carries a Bonn dateline of 9 October 2026. Observations concern September, not October.
• Bulletin data and analysis: method.html, text. ERA5 surface air temperature is 2m air temperature across all surfaces; anomalies are preferred to absolute temperatures; full global ranking record begins in 1940. Annual global pre-industrial conversion adds 0.88°C, while monthly offsets vary by month. This page does not specify whether multi-month surface-air-temperature averages give months equal or day-proportional weights.

Reproduce

From any working directory, run:

python3 stories/2026-temperature-record-src/calculate.py

The command checks the input hash, exact monthly calendar and schema, finite values, monthly climatology/offset consistency, and the two temperature identities (allowing source rounding for absolute minus climatology). It rewrites statistics.json using only the frozen CSV and the Python standard library. It does not contact the network. The JSON retains numerical precision for review; the outline plans two decimal places for visible temperature levels.

The main story arithmetic uses ano_91-20 only. Equal-month and calendar-day weighted alternatives are computed separately, including each convention's 2024 annual target and historical final quarters. This is a transparent conditional calculation on the supplied monthly anomalies, not a claim to reproduce C3S's unspecified annual aggregation or its eventual annual classification. No forecast, probability, physical uncertainty interval, or significance test is derived.