Map projections distort reality because a sphere cannot be flattened without stretching or tearing it, a fact Carl Friedrich Gauss proved in 1827. Every flat map therefore sacrifices at least one of four properties: area, shape, distance or direction. Conformal maps such as the Mercator keep local shapes and angles but inflate size away from the equator, so Greenland looks as large as Africa although Africa is about 14 times bigger. Equal-area maps such as Gall–Peters and Equal Earth keep every country's true size but warp outlines. Compromise maps such as the Winkel tripel, used by the National Geographic Society since 1998, accept small errors in everything to avoid a large error in anything. The clearest way to see what each choice costs is to set the projections against one another.
Gauss, 1827: Why No Flat Map Can Be True
Gauss's Theorema Egregium, Latin for "remarkable theorem", showed that a surface's curvature is intrinsic: it can be measured from distances and angles on the surface itself and does not change when the surface is bent without stretching. A sphere of radius R has curvature 1/R²; a sheet of paper has zero. Since no amount of bending changes curvature, no map can carry a patch of globe onto paper with every distance intact, however small the patch.
Cylinders, cones and planes are different. They are "developable" surfaces with zero curvature, so they can be unrolled flat without further damage. Most projections are therefore pictured in two steps: transfer the globe onto a cylinder, cone or plane, which is where the distortion happens, then unroll it. Peeling an orange and pressing the peel flat makes the same point by hand: it splits or stretches.
In 1859 and 1871 the French mathematician Nicolas Auguste Tissot gave cartographers a way to see the damage. Imagine tiny identical circles scattered over the globe and project them. Each becomes an ellipse whose size records area distortion and whose stretching records shape distortion. Those ellipses, called Tissot's indicatrices, are the common yardstick used below.
Mercator (1569) Against Gall–Peters: Angles or Areas
The sharpest contrast in cartography is between a conformal map and an equal-area one, because no projection can be both. Gerardus Mercator published his world map in 1569, on eighteen sheets measuring 202 by 124 cm, with a title announcing it was corrected "for the use of sailors". Its trick is to stretch north–south distances at each latitude by exactly as much as a cylinder stretches east–west ones. Tissot's circles stay circles everywhere, so angles are true and a line of constant compass bearing is straight. The price is that the circles swell toward the poles.
The Gall–Peters projection, described by James Gall in 1855 and relaunched by Arno Peters in 1973, makes the opposite choice. Its parallels crowd together toward the poles so that area stays constant; Tissot's ellipses all have the same area but are drawn tall and narrow at the equator and short and wide near the poles. Only along 45°N and 45°S is it free of shape distortion.
| Projection | Year | Designer | Keeps true | Gives up |
|---|---|---|---|---|
| Mercator | 1569 | Gerardus Mercator | Angles, local shape | Area, worse with latitude |
| Gall–Peters | 1855 / 1973 | James Gall, Arno Peters | Area | Shape, except at 45° |
| Winkel tripel | 1921 | Oswald Winkel | Nothing exactly | Small errors everywhere |
| Robinson | 1963 | Arthur H. Robinson | Nothing exactly | Polar shapes and sizes |
| Equal Earth | 2018 | Šavrič, Jenny, Patterson | Area | Shape, mildly in the tropics |
| Azimuthal equidistant | Described by al-Biruni, 11th century | Various | Distance and direction from one point | Area and shape away from the centre |
Greenland, Alaska and Madagascar: Mercator's Size Errors in Numbers
On a Mercator map, linear scale at latitude φ grows as the secant of φ, so area grows as its square. At 30° the area scale is about 1.33; at 60° it is 4; at 80° it is about 33. At the poles it becomes infinite, which is why a Mercator map can never show them and becomes practically unusable beyond about 70°.
Real places show what those numbers mean. Greenland, which runs from about 59°N to 83°N, is inflated roughly fourfold at its southern tip and more than 30-fold in the far north, so its 2,166,086 km² look comparable to Africa's 30.3 million km². In truth Greenland is about the size of the Democratic Republic of the Congo. Alaska looks as large as Australia, which is about 4.5 times bigger, and takes as much space as Brazil, which is nearly five times bigger. Madagascar and Great Britain look alike, yet Madagascar is more than twice as large. Australia itself, centred near 25°S, is inflated only by about 1.2; Great Britain, near 55°N, by about 3. On the Gall–Peters and Equal Earth maps all of these errors vanish.
None of this was hidden from Mercator. He never explained how he constructed his grid, and the English mathematician Edward Wright published the first accurate tables for it only in 1599. Sailors could not use it fully until the mid-18th century, when the marine chronometer made longitude reliable at sea. Mercator himself turned to the equal-area sinusoidal projection when he wanted to show relative sizes. The trouble began when publishers took a navigator's chart and hung it on classroom walls as a picture of the world; by the 1940s most atlases had dropped it for world maps.
Robinson (1963) Against Winkel Tripel (1921): The Compromise Projections
Most atlas world maps belong to neither camp. Arthur H. Robinson devised his projection in 1963 for the map publisher Rand McNally, working, as he put it, "backwards": he adjusted the outlines by eye until they looked right and only then derived the numbers. The result is defined by a table of values at 5° intervals rather than a formula. It is neither conformal nor equal-area, and it stretches each pole into a line. The National Geographic Society adopted it for world maps in 1988, replacing the Van der Grinten projection.
Ten years later it switched again, to the Winkel tripel, proposed by the German cartographer Oswald Winkel in 1921. Its name comes from the German for "triple": Winkel aimed to minimise three distortions at once, of area, direction and distance. It is the average of the equirectangular and Aitoff projections. When David M. Goldberg and J. Richard Gott measured world maps for distance error, ellipticity of Tissot's ellipses, area error and skew, the Winkel tripel fared better than the other projections they analysed and had the least skew of any. On another measure, Capek's "Q", it ranked ninth of a hundred, behind Robinson and Eckert IV. Different yardsticks crown different winners, which is itself the lesson.
Equal Earth (2018) Against Gall–Peters: Two Ways to Keep Area
Both maps show every country at true relative size, so the contrast lies entirely in shape. Gall–Peters uses straight meridians, which forces a shape distortion factor of two at the equator. Equal Earth, which Bojan Šavrič, Bernhard Jenny and Tom Patterson created in 2018 in direct response to Boston Public Schools' 2017 adoption of Gall–Peters, curves its meridians like the Robinson. Shapes are exact on the central meridian at 40.4° north and south; nearer the equator they are elongated north–south by at most a factor of 1.35. Poleward of 40.4° the distortion grows again and reaches extremes near the poles, which are drawn as lines.
Its equations were designed to be simple and fast to compute, and the open-source PROJ library added it in September 2018. Eight years later it won a UN endorsement: on 4 September 2026 the General Assembly voted 164 to one, with six abstentions, to discourage Mercator for most uses other than navigation and to encourage equal-area maps, naming Equal Earth as an option.
Distance and Direction: The UN Emblem's Azimuthal Equidistant Map
Area and shape are not the only properties at stake. An azimuthal equidistant projection keeps both distance and direction true, but only from its centre point. The United Nations emblem, adopted on 7 December 1946, uses a polar version centred on the North Pole and reaching 60°S: every meridian is a straight spoke and every distance from the pole is correct. Distances between two other points, such as Lagos and Lima, are not, and the far southern lands are smeared into a ring. Centre the same projection on Tokyo and it gives correct great-circle distances and bearings from Tokyo to everywhere, which is why radio operators use maps centred on their own antenna to find the bearing to a distant station.
Web Mercator Since 2005: The Distortion on Every Phone
The most widely viewed projection today is a variant of the oldest one discussed here. When Google Maps adopted Web Mercator in 2005 it became the standard for online maps, now used by Bing Maps, OpenStreetMap, Mapbox, Esri and others. For street navigation it is ideal, because conformality means a city block is never sheared and north is always up. Zoom out, though, and the 1569 problem returns: Greenland swells, and the map is simply cut off at 85.051129° north and south, the latitude at which the whole world becomes a square. The US National Geospatial-Intelligence Agency considers it unacceptable for official use, because treating its coordinates as true Mercator can shift positions by as much as 40 km.
For a deeper look at the designs compared here, see our guides to the Mercator projection, the Winkel tripel and the Robinson projection, or follow the equal-area argument in the Peters projection explained.
