Frozen Pipes in Washington

A single state number would hide the thing that decides this: elevation, distance from water and whether the local terrain traps cold air move these figures further inside Washington than a whole decade of warming does. Each row is one NOAA reference station, and every figure in the table is our own count from that station's daily minimum and maximum temperatures. NOAA publishes threshold day counts at 32 F and at 0 F and nothing at 20 F, so none of this is a quoted normal.

Everything below is read live from api.weather.gov when this page loads and shown exactly as the issuing forecast office published it, with the office named. Each column in the seven night strip is one NIGHT, labelled for the evening it starts on in local time.

A wind chill of 25 below sounds like an emergency for plumbing, and in one narrow sense it is: wind strips heat off a cold pipe faster, so the pipe reaches the air temperature sooner. What it does not do is take the pipe below the air temperature. The National Weather Service says this plainly, and it is the single most useful sentence published about this subject:

So the number to watch for a building is the forecast air temperature, and the number to watch for a person outside is the wind chill. They answer different questions. The seven night strip above this section shows air temperature for exactly that reason.

And the product names changed. On 1 October 2024 the National Weather Service retired five cold product names and replaced them. This matters more than a naming tidy up: any site or app still querying a retired name gets an empty result from the agency's feed, and an empty result renders as no alerts, which looks exactly like an all clear.

One consequence worth knowing: the National Weather Service's own cold weather FAQ page still described advisory criteria under the old wind chill names when we read it in August 2026. Even the agency has pages that have not caught up, which is why the retired names are still everywhere.

Derived from NOAA daily records. Nothing in the Washington table is a published NOAA normal, because NOAA publishes none for these measures: its 1991-2020 normals give threshold day counts at 32 F and at 0 F, no date shaped cold variable, and nothing about how long a cold spell lasts. Every count here is computed from NOAA NCEI GHCN-Daily daily minimum and maximum temperatures over the 30 winters from 1991-92 to 2020-21 and committed as a dated snapshot.

The rules are the same at every station. A hard freeze night is a night whose minimum reached 20 F or colder. An ice day is a day whose maximum never rose above 32 F. A winter runs July to June and counts only if at least 90 percent of its November through March nights reported; incomplete winters are excluded rather than filled in. Percentile dates are the 10th, 50th and 90th over the winters that qualified, and where the median lands on a winter with no hard freeze at all, no date is printed. Records and worst events use each station's full daily record, which runs back to 1991.

One data quality note. NOAA's daily summaries product carries a small number of unflagged impossible values, and three stations in this dataset have one: Mobile reports 41 below zero on 28 June 1999, Orlando reports 1 F on 31 July 2009, and Biloxi reports 0 F on 11 September 2002. All three are obviously instrument or transcription failures, and at stations where a hard freeze is rare a single bad value takes over the whole event ranking. Any hard freeze record or event dated June through September is dropped. That filter removes exactly those three values across all 148 stations, and it costs nothing at the high elevation stations whose ranked events all fall in deep winter.

What the record counts. It counts weather: no agency collects burst pipe counts for Washington or anywhere else, and a station's record describes that station rather than any particular building. The page's shape follows its numbers: where a state's primary station averages 30 or more hard freeze nights a year the page leads with stretch length, where it gets there in fewer than 50 percent of winters the page leads with the record rather than a median date, and Washington is on the rare shape.

Live conditions come from api.weather.gov, read in your browser at page load. Source: NOAA NCEI GHCN-Daily, a US Government work in the public domain. The February 2021 loss figures elsewhere on this site come from the Texas Department of Insurance loss compilation, which counted 510,772 claims for a single nine day event.