Fifty runways are the nearest tarmac for half the world
Take every runway on Earth a widebody could physically use — paved, open, at least 8,000 feet — and colour every point on the planet by which one is nearest. There are 2,280 of them. Fifty are the answer for nearly half the world.
A Voronoi on a sphere, not on a map
The usual way to do this is to project the world flat and compute the diagram in map coordinates. That is wrong, and visibly so near the poles and the date line, where flat distance stops corresponding to real distance. This is computed as a spherical Voronoi — a Delaunay triangulation on the sphere, obtained as a 3D convex hull, with circumcentres for vertices — so every boundary is a genuine "closer to A than to B" line on the actual planet.
It is then drawn on an equal-area projection, because the chart's claim is about area. Telling an area story in Mercator is a lie: it would inflate the polar cells that already dominate and make the headline look bigger than it is.
The 50 largest cells cover 48.7% of the Earth's 510 million km². The continents dissolve into a dark mosaic of thousands of tiny cells, while a handful of ocean strips swell into the biggest shapes on the map. Mataveri on Easter Island is the nearest usable runway for 16.4 million km² — 4,624 times the ground Heathrow is nearest to, at 3,557 km².
The list of giants is largely remote-island and military: Wake, Ascension, Mount Pleasant, Diego Garcia, and Hao, a 1960s French nuclear-test support strip. The pattern is not about traffic. It is about emptiness.
The check that caught a real error
An early version had the Pituffik cell inside out — the polygon that wraps the North Pole. Our own spherical-area formula returned the whole sphere for it, which is the classic failure mode when a cell contains a pole: the winding order flips and "inside" becomes "outside". The fix was to stop computing it ourselves and ask the library that draws it, so the number and the picture come from the same source.
A second check mattered as much. Our sampler used equal-area latitude bands but counted each sample equally, which under-weighted the high-latitude giants — the ones that carry the headline. Area-weighting moved the answer from 47.1% to 48.7%, matching the exact polygons.
What the chart does not claim
"Nearest" here is great-circle distance to a runway that meets the physical criteria. It says nothing about whether an aircraft would be permitted to land, whether the airport is staffed, whether it has fuel, or whether anyone would divert there. It is a geometry of tarmac, not a map of operational alternates — which is stated on the chart, because the difference matters to anyone in aviation.
What your data needs to look like
A point and an identifier. That is the whole requirement.
ident
name
lat
lon
SCIP
Mataveri International Airport
-27.165
-109.421
EGLL
London Heathrow
51.4706
-0.4619
Novice tip: any "which of these is nearest?" question over a set of points is this chart — retail catchments, hospitals, chargers, transmitters. If the points span more than a few hundred miles, insist the diagram is computed on the sphere.
The takeaway
Infrastructure maps usually show where things are, which is a map of population. Colouring by nearest instead shows where things are not, and the result is a picture of the planet's emptiness that a dot map cannot produce.