Infrastructure

Which way runways point

—

Tap a region to jump to it.

Every runway OurAirports records with two separately surveyed ends — 9,241 of them, grouped into thirteen regions — drawn on one dial at its true compass bearing, so every stroke lies parallel to the real strip of tarmac or grass. Distance from the middle is the runway's length. Press play to run the regions from the most aligned to the least. Where the wind is steady, the runways obey it: 45% of Caribbean runways lie within 20° of 080°, square into the trade winds, against the 22% an even spread would give, and Northern Europe leans the same way under the westerlies. Where the wind is variable the dial just fills in — Australia manages 25%, which is nothing. Then the American Midwest breaks it: 22% sit within 20° of its own strongest axis, no better than chance, yet 34% sit within 7.5° of true north or true east — twice an even spread — on land that was surveyed into a mile grid before it had airfields. Tap or hover any runway.

Data & source

Data: OurAirports — runways.csv and airports.csv, released by its contributors into the public domain · fetched —.

How the bearing is measured. Not from the file's heading_degT columns, which are partly derived — 2,019 of them sit on an exact multiple of ten, which is the runway number (runway 18 → 180°) written back into the heading field rather than a survey. Each bearing here is the great-circle bearing between the runway's two surveyed threshold coordinates, which is a measurement of where the strip physically lies. Where both exist the two agree closely (median 0.25°, 90th percentile 1.4°). 611 rows whose geometric bearing landed on an exact integer were dropped: those are ends back-computed from the runway number, and keeping them would manufacture the very cardinal spike this post is about. A runway is bidirectional — 080°/260° is one strip, not two — so bearings are folded onto 0–180° and each is mirrored across the dial. Nothing is double-counted.

Scope. OurAirports is community-maintained and its coverage is uneven: only about a third of its runway rows carry two surveyed thresholds, and the United States is far better covered than anywhere else — which is why the regions here are compared with each other and never summed into a world total. These are the runways that are mapped, not the world's runways. Heliports, seaplane bases and runways marked closed are excluded. Of 14,170 surveyed runways worldwide, 9,241 fall inside the thirteen regions below; the rest are outside every box. Two rows carry no length and are drawn at the innermost radius.

Alignment is the resultant length of the doubled bearings — the standard statistic for axial data, where 010° and 190° are the same strip. 0 means every direction is equally used; 1 means every runway in the region is parallel. Mean axis is the direction those doubled bearings point on average, and it is only meaningful when alignment is high; for a region near 0 it is close to noise, which is why the dial draws it fainter the weaker it gets. An even spread would put 8.3% of runways within 7.5° of north and another 8.3% within 7.5° of east.

What the data does and does not say. That runways in the trade-wind and westerly belts line up, and that Midwestern runways cluster on true north and true east, are both measured here. Why is not: this file records no wind and no survey history. Aircraft take off and land into wind, and the American interior was subdivided by the Public Land Survey System into north–south, east–west mile sections long before it had airfields — those are the standard explanations for the two patterns, and they are offered as explanations, not as findings. Regional boxes were chosen by the pipeline and are printed in full in the table; the three American boxes do not overlap.