What Is a Weather Record and How Is One Set?

A weather record is the highest or lowest value a weather station has measured for a given element, on a given calendar day, in a given month, or ever. It is a statement about one place and one measurement over the years that station has been observing, not a statement about the planet.

The parts of a record

Every record has the same parts. Knowing them makes any record headline easier to read.

The station. A record belongs to a station, a fixed site with instruments. A city can have a record only through a station, or through several stations joined into a thread when a station has moved.

The element. We track five measurements from the GHCN-Daily dataset: daily maximum temperature, daily minimum temperature, precipitation, snowfall and snow depth. They give seven record types: highest maximum, lowest maximum, highest minimum, lowest minimum, highest daily precipitation, highest daily snowfall and greatest snow depth.

The timescale. A daily record is the extreme for that calendar day across all the years on file, for example every 14 July. A monthly record is the extreme for any day of that month. An all-time record is the extreme for the station ever. One observation can break all three, and we count it once, under the broadest timescale it breaks.

The value and the record it replaced. A new record has a value and a previous value with its date. The gap is the margin, always shown as a positive amount, and the age of the old record is the number of years it stood.

The period of record. This is how long the station has been observing. A record from a station with 120 years of data means more than one from a station with 31 years.

How a record is set

Each day, every observation is compared with the standing values for that station, element and calendar day. If a new value is higher than the highest one (or lower than the lowest one, for the low types), a record is set. If it equals the standing value at the recorded precision, it is a tie, which has its own explainer. Precipitation, snowfall and snow depth have only a high side, since zero is the floor.

Leap days are handled explicitly. 29 February has its own daily record, built only from leap years, and it is tested as a case of its own.

Which stations can set records

A station has to qualify. It needs at least 30 years in which more than 182 days have a complete observation for the element in question. That test is applied as of each date, not as of today, so a station begins to produce records once it has qualified and not before. The rule keeps very young or patchy stations from producing records that mean little. It also matches the approach used by NOAA’s own records tool, so comparisons are fair.

Why raw counts mislead

The number of records set on a day depends on how many stations qualify and how old they are. More stations means more records. A longer history means a harder bar. So a total of records is a poor way to compare years or countries. The ratio of record highs to record lows sidesteps most of this, because both counts come from the same stations under the same rules. If the climate were stable and stations were long lived, the ratio would sit near one to one.

Measurement limits

Records are raw observations, not adjusted for anything. Several things can affect them:

• Instrument and site changes. A new sensor or a station moved to a different spot can shift readings. City pages state where a thread joins stations.
• Reporting practice. Observation times differ between countries, and this can shift which calendar day a value falls on.
• Missing days. A missing observation is not zero. Only complete observations are compared, and flagged values are dropped.
• Quality control after the fact. A value can be corrected weeks later, which is why recent records are marked provisional.

Reading a record page

On any record on this site, read four things in order. First the value and the unit, shown in Celsius and millimetres with Fahrenheit and inches beside them. Second the margin over the old record. Third the years the old record stood. Fourth the significance score from 0 to 100, which combines margin, age and period of record so a small daily record is not ranked with a big all-time one. The rule is set out on the methodology page, and the age of records across the network is on the longevity page.

What a record does not tell you

A record says that a value was more extreme than any before it at that station. It does not say the weather was dangerous or that it was caused by any one thing. Many records in one week across many stations, however, is evidence of a large-scale pattern, which is how we build heat waves and other events from the data. For the observation source and its licence, see the sources page.