Methodology
Every number on this site traces to a stored observation from a named source. Nothing is modelled, estimated or inferred. This page is generated from the same source register the ingest uses, so it cannot drift from what is actually being collected.
Where the readings come from
Two independent networks supply air temperature. METAR is the aviation network: airports worldwide, reporting continuously, roughly every 30 to 60 minutes. SYNOP is the national weather service network, reporting at the main synoptic hours of 00, 06, 12 and 18 UTC.
Both are needed, and the reason is specific. METAR is an airport network, and the coldest places on Earth have no airport. Measured on 22 September 2026, the coldest METAR station on the planet was McMurdo at −20.1 °C, with positions three to ten filled by ordinary airports between −3 and −4 °C. SYNOP for the same moment carried Vostok at −70.4 °C. Without SYNOP the cold half of this site would be wrong by roughly 50 degrees while still looking entirely plausible.
Where a station appears in both networks, both readings are kept. That is useful rather than redundant: it gives two independent measurements of the same place in two different formats, which is one of the ways a reading gets confirmed.
Quality control
A reading cannot be shown, ranked or recorded unless it passes every rule:
- Not older than its network allows: 2 hours for METAR, 9 hours for SYNOP.
- Within the absolute range of −95 to 60 °C.
- Not more than 10 °C different from the same station an hour earlier.
- Not more than 12 °C from the median of stations within 150 km, when at least three such stations exist.
- Has a real position. These feeds use placeholder coordinates for unknown locations, and a reading with no location is discarded rather than placed in the ocean.
- Not manually suppressed.
The staleness rule differs by network because their cadences differ. A single 2 hour rule applied to SYNOP would discard the entire network, including every Antarctic station. The observation time is shown beside every reading instead, so a six hour old report is visibly a six hour old report.
How the hottest and coldest are decided
Passing quality control is not enough to be called the hottest or coldest place on Earth. A headline reading also needs a second, independent signal: either another reading from the same station agreeing within 1.5 °C, or a station within 150 km agreeing within 5 °C.
A reading that passes quality control but has nothing to corroborate it is marked unconfirmed. It stays in the table, because removing it would mean the table no longer shows the most extreme reading we actually have, but it cannot be the headline figure, set a record or be posted. A station reporting for the very first time is always held for one further cycle.
This is why the headline number is sometimes not the number at the top of the table. When that happens the page says so and shows both.
Air temperature and ground temperature
Station readings are air temperature, measured about 1.5 to 2 m above the ground. Satellite readings are land surface temperature, the temperature of the ground itself, which in a daytime desert runs far hotter than the air above it.
These are different quantities. They are shown together, because that is what readers want, but they are never ranked against each other, never combined into one list and never share a record. Satellite values are pinned above the table without a rank number and always carry a ground label.
Sources in use
| Source | Measures | Ranked | Licence |
|---|---|---|---|
| NOAA Aviation Weather Center, bulk METAR cache METAR, airport station | Air | Yes | US Government work, public domain |
| NOAA NWS raw WMO SYNOP bulletins SYNOP, WMO station | Air | Yes | US Government work, public domain |
| WMO WIS2 Global Cache SYNOP via WIS2 | Air | No | WMO Unified Data Policy core data, free and unrestricted, attribution requested |
| NASA LANCE VIIRS land surface temperature, near real time VIIRS satellite | Ground | Yes | NASA, public domain |
| NOAA GOES-R ABI land surface temperature GOES satellite | Ground | Yes | US Government work, public domain |
| NOAA GHCN-Daily GHCN-Daily | Air | No | US Government work, public domain |
| NOAA GHCN-hourly GHCNh | Air | No | CC0-1.0 public domain dedication |
| Meteostat bulk Meteostat | Air | No | CC BY 4.0, commercial use permitted |
Sources deliberately not used
Some otherwise useful sources are excluded on licence or terms. They are listed so the decision is visible rather than implied.
- Ogimetrobots.txt is "User-agent: * / Disallow: /", a blanket ban on automated fetching.
- Open-Meteo free APITerms permit the free API for non-commercial purposes only and name ad-supported sites as commercial. This site carries ads.
- JAXA Himawari Monitor (P-Tree)Non-commercial terms and redistribution prohibited.
- MODIS Terra and Aqua, live useTerra and Aqua have drifted from their original orbits and NASA is transitioning to VIIRS. Historical context only.
- Crowd-sourced personal weather stationsRooftop and sun-exposed siting produces false extremes.
Attribution
- NOAA Aviation Weather Center (US Government work, public domain)
- NOAA National Weather Service, and the reporting national meteorological services (US Government work, public domain)
- WMO WIS2, and the reporting national meteorological services (WMO Unified Data Policy core data, free and unrestricted, attribution requested)
- NASA LANCE VIIRS (NASA, public domain)
- NOAA GOES-R (US Government work, public domain)
- NOAA NCEI GHCN-Daily (US Government work, public domain)
- NOAA NCEI GHCNh (CC0-1.0 public domain dedication)
- Source: Meteostat and its data providers (CC BY 4.0, commercial use permitted)