The Forgotten Art of Map Projection

You cannot wrap a sheet of paper around an orange without crumpling it. This is not a metaphor — it is a theorem, proved by Carl Friedrich Gauss in 1827 and called theorema egregium, the "remarkable theorem." A sphere has intrinsic curvature; a plane does not. Therefore any flat map of a curved surface must distort something. Distance, angle, area, shape: pick at most two to preserve. The history of cartography is the history of choosing which lies are tolerable.

Mercator and the navigator's bargain

Gerardus Mercator published his world map in 1569, and within a generation it was the standard for ocean navigation. The reason is mechanical: on a Mercator projection, a line of constant compass bearing — a rhumb line — appears as a straight line on the chart. A captain in the Atlantic could draw a ruled line from Lisbon to Recife, read the bearing once, and steer it for weeks. The math behind this convenience is brutal. To preserve angles everywhere, Mercator stretches the map vertically toward the poles, by a factor of one over the cosine of latitude. At 60° N this is a doubling. At 80° N it is a sixfold inflation. At the poles it is infinite.

The familiar consequences: Greenland appears the size of Africa, though Africa is fourteen times larger. Antarctica appears as an unbroken ice rim that wraps the bottom of the world. Russia and Canada loom continental beyond their actual mass. None of these were lies Mercator told for political reasons; they were the price of the rhumb line. He was making a tool for sailors, and he was honest about what he had bought and what he had sold.

The Peters projection and the political reframing

In 1973 Arno Peters, a German historian and self-taught cartographer, presented an "equal-area" world map and accused the entire cartographic establishment of centuries of Eurocentric distortion. His map preserved area — Africa really is fourteen times Greenland — but at the cost of grotesque shape distortion: continents stretched into vertical streaks, the equator wider than the poles by a factor of pi.

The professional response was withering. The American Cartographic Association noted, accurately, that Peters had reinvented an 1855 projection by James Gall and was selling it as new. They also pointed out that he had treated cartography as a binary moral choice when in fact dozens of equal-area projections had existed for centuries (Mollweide, Hammer, Eckert, Goode), most of them more elegant than his.

And yet Peters won the cultural argument he wanted to win. His map showed up in classrooms, in UNDP publications, in Boston Public Schools' official curriculum in 2017. He had reframed projection as politics — and he was right that it always had been, even when no one was talking about it. Mercator's map, by virtue of being the default, had been a political act for four hundred years; Peters just made the politics legible.

Robinson and the compromise of accepting compromise

By the late 20th century the National Geographic Society needed a world map that didn't lie spectacularly in any one direction. Arthur H. Robinson, professor at the University of Wisconsin, designed one in 1963 by reverse engineering aesthetic intuition: he plotted what continents should "look like," tabulated the coordinates, and then derived the math. It is, formally, a non-equal-area, non-conformal pseudocylindrical projection. Informally: it just looks right. National Geographic adopted it in 1988, kept it as their default for ten years, then replaced it with the Winkel Tripel in 1998 — another compromise, slightly more accurate at the poles.

The Robinson is honest in a different way than Mercator or Peters: it admits that for a general-purpose world map, no property matters enough to preserve perfectly, and the goal is to keep all the lies small. It distorts area by up to 25 percent, distance by similar margins, and angles everywhere — but never catastrophically. This is the cartographic equivalent of a generalist database: not the best at any single query but never embarrassingly bad at any of them.

The web Mercator complication

When Google Maps shipped in 2005 it used a variant of Mercator (technically WGS 84 / Pseudo-Mercator, EPSG:3857), and the choice has shaped a generation of digital map design. The reason was again mechanical: Mercator preserves local shape and angles, which means at the city level — the scale where people use slippy maps — streets meet at the right angles you expect from looking out the window. A car-navigation overlay on an equal-area projection would warp intersections at high latitudes into impossible kinks.

The cost is the old one. At zoom level zero, looking at the whole earth, web Mercator inherits all of Mercator's distortions. Greenland still towers. The Arctic Ocean appears more than half the planet. Most users never notice because they never zoom out that far — the projection's lies are hidden by the fact that the relevant lies happen at scales no one looks at.

What every projection knows

The deepest thing the projection wars teach is that every flattening is an editorial act. A map that preserves area cannot preserve shape. A map that preserves angle cannot preserve area. A map that compromises everywhere preserves nothing perfectly. There is no neutral ground — even the "default" view in your map app is making aesthetic and political claims about what matters.

This generalises far beyond cartography. Every model of a complex system — a ER diagram of a database, a system architecture poster on the wall, an org chart, a financial dashboard — is a projection. It chooses what to preserve and what to flatten. The honest practitioner names the trade-off explicitly: this picture preserves cause and effect at the cost of timing; this one preserves quantity at the cost of relationship; this one preserves intuition at the cost of numerical accuracy.

Mercator did not have to defend his projection because his readers were sailors who knew exactly what it cost. Today's defaults are inherited rather than chosen, and most users have no idea what's been distorted to give them the view they're looking at. The forgotten art of projection is partly geometric — but it is mostly the discipline of remembering that the picture is not the thing, and that the choice of picture is itself a claim.