Map Projections Explained: Mercator, Web Mercator, and the Trade-offs

The Earth is round. Maps are flat. That mismatch is the entire problem of cartography, and a map projection is the mathematical solution. Every projection that exists is a deliberate compromise: pick which kind of distortion hurts your use case least, and accept the rest.

This guide explains what a map projection is, the main families, why every web map uses Web Mercator, and which projection to pick when accuracy actually matters.

Why Projections Exist

A globe is the only honest representation of the Earth. The moment you flatten the sphere onto a 2D surface, you have to stretch, tear, or compress it somewhere. There is a famous mathematical result, Gauss's Theorema Egregium, that proves there is no way to flatten a sphere without distorting distances. So instead of asking "what is the perfect projection?", cartographers ask "which distortion can I afford?"

Every projection trades among four properties:

• Shape (conformality): angles and local shapes are preserved
• Area (equivalence): regions are shown at their true relative sizes
• Distance (equidistance): distances from one or two points are accurate
• Direction (azimuthality): directions from one point are accurate

No projection preserves all four. Most preserve one and distort the rest in controlled ways.

The Mercator Projection

Gerardus Mercator published his projection in 1569. It is a cylindrical projection that wraps the Earth in a cylinder tangent to the equator and unrolls it. The key property is conformality: at any point on the map, angles are preserved, which means a constant compass bearing (a rhumb line) is a straight line on the map. That made Mercator perfect for ship navigation. A captain could draw a straight line, read a bearing, and sail it.

The cost is brutal area distortion. The further you go from the equator, the more the projection stretches features. Greenland appears the size of Africa (Africa is 14 times larger). Antarctica appears as a giant blob along the bottom edge. Russia and Canada look enormous compared to equatorial countries.

This is why Mercator is criticised in education and journalism: it warps the mental model of how big countries actually are. Several alternative projections, like Gall-Peters and Equal Earth, were designed specifically to fix this for general-audience world maps.

Web Mercator and Why It Won

In 2005, Google launched Google Maps with a slightly modified Mercator projection now called Web Mercator (EPSG:3857). Web Mercator differs from classic Mercator in one practical way: it treats the Earth as a perfect sphere instead of the more accurate ellipsoid (WGS84). That makes the math faster and lets tile servers cut the world into a clean pyramid of square tiles.

Web Mercator dominates the web for one operational reason: every web map can share the same global tile grid. Zoom level 0 is one tile covering the whole world. Zoom level 1 is four tiles. Zoom level n is 4^n tiles. Panning and zooming are reduced to fetching a different set of small image or vector tiles. The whole slippy map experience that powers Google Maps, Apple Maps, OpenStreetMap, MapAtlas, and every embedded web map you have ever used is built on this assumption.

The math, in one paragraph

For longitude lng and latitude lat in radians, Web Mercator computes:

• x = R * lng
• y = R * ln(tan(pi/4 + lat/2))

where R is 6,378,137 metres, the equatorial radius of WGS84, applied to a sphere. At zoom 0 the whole world fits in one 256 by 256 pixel tile, and each zoom level doubles the resolution in both directions.

The ln(tan(...)) term is why areas stretch toward the poles. The spacing between parallels grows with latitude, and the local scale factor is 1/cos(lat). At 60 degrees north, roughly the latitude of Helsinki or Oslo, distances are drawn about twice as long as at the equator and areas about four times as large. The same formula goes to infinity at the poles, which is why the projection stops at about 85.05 degrees and why polar regions are clipped on every web map world view.

The trade-offs of Web Mercator are real but acceptable for most product use cases:

• Same area distortion as Mercator (Greenland still looks huge)
• Cuts off the poles around 85.05 degrees latitude (because the math goes to infinity at the poles)
• Slight inaccuracy versus a true ellipsoid model

For a navigation map, a real estate map, or a delivery zone map, none of those matter. For analysis of the polar regions, choropleth comparisons of country sizes, or anything where area matters, Web Mercator is the wrong tool.

Other Major Projection Families

Equal-area projections

Equal-area (equivalent) projections preserve the relative size of regions. Shapes get distorted instead. They are the right choice for any thematic map where the reader will compare areas (population density, land use, election outcomes by country).

Common equal-area projections:

• Mollweide: oval-shaped world map, good for global thematic visualisations
• Equal Earth: a 2018 projection designed to look natural while preserving area
• Albers Conic: standard for the contiguous United States and similar mid-latitude regions
• Gall-Peters: famous for being explicitly anti-Mercator in education contexts

Conformal projections

Conformal projections preserve angles and local shapes. Mercator is the most famous, but there are many others tuned for specific regions:

• Lambert Conformal Conic: standard for aeronautical charts and US state plane coordinate systems
• Stereographic: preserves circles, used for polar regions and small-area maps

Equidistant projections

Equidistant projections preserve distances along specific lines. The Azimuthal Equidistant projection, when centred on a point, shows all distances and directions from that point accurately. It is the projection used on the United Nations flag (centred on the North Pole).

Compromise projections

Compromise projections do not preserve any single property exactly but balance distortions to look natural. They are popular for general reference maps.

• Robinson: from 1963, National Geographic's world map projection from 1988 to 1998
• Winkel Tripel: adopted by National Geographic for world maps in 1998
• Natural Earth: a 2008 compromise by Tom Patterson, designed for world maps of physical data

Choosing a Projection

Start with the question.

• Web map for product or navigation: Web Mercator, no debate
• World thematic map (population, climate, election): Equal Earth or Mollweide
• Single-country or continental thematic map: pick an equal-area projection optimised for the region's latitude and shape (Albers for the US, Lambert Azimuthal Equal-Area for Europe, etc.)
• Polar region: Stereographic or Azimuthal Equidistant
• Navigation chart: Mercator or Lambert Conformal Conic
• General-purpose world reference map: Winkel Tripel or Robinson

A good shortcut: if you are doing analysis where area matters, use an equal-area projection. If you are putting an interactive map on a web page, use Web Mercator and stop worrying about it.

Coordinate Systems and EPSG Codes

Every projection is identified by an EPSG code. The most common in the wild:

• EPSG:4326: WGS84 lat/lng, the default for GPS and most data exchange (not actually a projection, just lat/lng on the ellipsoid)
• EPSG:3857: Web Mercator, used by virtually every web map
• EPSG:3035: ETRS89 / LAEA Europe, equal-area projection for European thematic maps
• EPSG:5070: NAD83 / Conus Albers, equal-area for the contiguous United States

When data lands on your desk in EPSG:4326 (lat/lng) and you need to display it on a web map, the tile server reprojects it to EPSG:3857 on the fly. When you need to do area calculations, reproject to a local equal-area projection first.

In practice, keep 4326 as the source of truth everywhere except the rendering boundary: PostGIS columns as geometry(Point, 4326), JSON payloads as [lng, lat] in WGS84, and map library input in 4326 that the renderer projects internally. Storing data in 3857 is almost always a mistake. The moment you compute a distance, send the data to another provider, or run it through a GIS tool, you have to convert back, and each round trip costs precision.

Working with Projections in MapAtlas

The MapAtlas Maps API serves vector and raster tiles in Web Mercator (EPSG:3857), the same as every other web map provider. When you load a base style and add a GeoJSON layer in EPSG:4326, the renderer handles the reprojection for display. For coordinate work, the coordinates lookup tool returns standard WGS84 lat/lng (EPSG:4326).

For analysis where you need an equal-area projection (true area calculations, accurate centroids, density per square kilometre), do the reprojection in your data pipeline before the data hits the map. Tools like PostGIS, GDAL, or Turf.js handle the conversion between EPSG codes.

A map projection is a tool for thinking, not a feature of reality. Pick the one that matches your question and accept what it asks you to give up in return.