Why runway numbers drift the pavement never moves; north does

A runway's number is its magnetic bearing, rounded to the nearest ten degrees and divided by ten. Runway 27 points roughly 270° magnetic. That definition ties a painted number to the Earth's magnetic field — and the field drifts, a few hundredths of a degree a year, forever. The concrete stays exactly where it was poured, and the number slowly stops describing it.

This page checks that, rather than asserting it. Every numbered runway end in the United States, taken from the FAA's NASR survey, is converted from its true alignment to a magnetic one using the World Magnetic Model, and asked a single narrow question:

If this runway were numbered today, by the rule, would it get the number it currently wears?

For 4,392 of 15,999 runway ends — 27% of them — the answer is no. 2,656 of those are paved, and 542 are at towered fields.

How a number is earned, and how it is lost

Rounding to ten degrees means a correctly numbered runway is already up to 5° away from its designator on the day it is painted: a runway on a bearing of 094° and one on 086° are both runway 09. That 5° band is the budget. Drift spends it. Once the magnetic bearing passes the halfway line the rounding rule names a different runway, and the number on the pavement is, strictly, wrong.

Being wrong is not the same as being repainted. Renumbering a runway means new paint, new signage, new charts, new instrument approach plates, new ATC phraseology and a NOTAM campaign, so airports let the discrepancy sit for years and change the number when something else was going to be repainted anyway. The 4,392 ends below are not an error log — most are simply living with it.

How predictable is the drift?

Over a few years, very. The field's rate of change is smooth and the model carries it explicitly, so a five-year projection is arithmetic. Over decades it is not predictable at all: the rate itself changes, in steps the models cannot see coming — geomagnetic jerks, sudden shifts in the field's acceleration originating in the core, which arrive every few years and are only ever identified afterwards. That is exactly why the WMM is reissued every five years instead of being solved once.

Direction matters as much as speed. Declination is not increasing everywhere — across much of the country it is falling, which walks those runway numbers back toward correct. Drift is not a ratchet.

Where it is going

Staying inside the model's validity window, 444 runway ends that are correctly numbered today will be outside the 5° band by 2030:

• 102 more ends outside the band within 1 year
• 210 more ends outside the band within 2 years
• 444 more ends outside the band within 4 years

The queue is long and the front of it is crowded — thousands of ends sit inside the band, so the ones nearest the line are separated by months. These are the paved, public-use runways closest to crossing:

The high score

The runways furthest past the line: paved, public-use fields where the painted number and the rounding rule disagree by the widest margin. Each row is one strip of pavement — both ends of a runway drift together and by exactly the same amount, so listing 09 and 27 separately would only say it twice.

What is deliberately not in that table

Three things look like drift and are not, so they are classified out rather than allowed to win a ranking they do not belong in.

• Reversed ends (20). NASR has the two ends of these runways swapped, so an end sits 180° from its own number. That is a data error, not magnetism, and it would otherwise own the top of this table permanently.
• Deliberate parallels (119). Some airports number near-parallel runways apart on purpose to keep them distinguishable on frequency — Denver's 7/25 lies about a degree from its own 8/26. The rule yields to operations, correctly.
• Never precisely numbered (807). Ends further off than declination could have carried them in ninety years at that location. Almost all are private turf strips whose number was eyeballed, not surveyed. The ceiling is computed per location, because eastern Alaska drifts about seven times faster than the national median and a gap that is impossible in Ohio is unremarkable at Barrow.

The reversed ones, for the record

Published rather than quietly dropped — they are a real finding about the source data, and this site already keeps a public ledger of what it gets wrong.

Method & sources

• Alignment: FAA NASR 28-day subscription, true headings as surveyed. 16,945 numbered runway ends at 6,459 US airports. Helipads, compass-point strips and seaplane lanes are excluded — only a numbered runway makes a claim about a magnetic bearing.
• Declination: WMM2025 (NOAA/NGA/BGS, public domain), evaluated here to degree 12 and checked against NOAA's own published test values to better than 0.01°.
• Not NASR's magnetic variation. NASR carries a whole-degree variation per airport, but it is a value of record carried forward from the last survey, and it disagrees with the model by a median of 2.0° — up to 14° at the worst field. Using it would measure how stale the FAA's paperwork is rather than where north actually points.
• “Would be numbered differently” is not “has drifted.” NASR does not record when a number was painted, so the cause is never claimed — only the present-tense comparison, which the data settles outright.

Advisory only, and emphatically not for navigation. The number painted on the runway is the number ATC will use and the number you will read back. Nothing here changes that.

Recomputed from the live NASR table; last generated 2026-10-09 07:30 UTC. See also how the site works and the airport directory.