Maps today are made by assembling digital layers in a geographic information system (GIS). Satellites such as Landsat (in orbit since 1972) and Europe’s Sentinel-2 (10 m pixels) photograph the surface; the 24-satellite GPS constellation and the WGS 84 coordinate system pin every feature to a position; radar from the 2000 Shuttle Radar Topography Mission and aircraft-mounted LiDAR measure the height of the ground; camera cars like Google’s Street View, launched in 2007, record what is at street level; and volunteers on OpenStreetMap, founded in 2004 and past one million registered users by 2013, trace roads and buildings by hand. Cartographers then decide what to keep, simplify and label for each scale, and publish the result as a zoomable online map that can be updated within days of a change on the ground.
That pipeline replaced centuries of chain, theodolite and pen. Great Britain’s Ordnance Survey, set up to map Scotland after the Jacobite rising of 1745, now earns only about 5% of its annual revenue from paper maps; the rest comes from digital data, route planning and apps. Since 1 April 2015 it has operated as Ordnance Survey Ltd, a company wholly in public ownership, producing digital data for business and government alongside its paper maps. The steps below follow the order in which a modern map is actually built.
Fixing a Position: WGS 84 and the GPS Constellation Since 1993
Every map starts with a coordinate system, because a road traced from a photograph is useless until each point has an agreed latitude and longitude. The usual reference is WGS 84, the World Geodetic System of 1984, maintained by the US National Geospatial-Intelligence Agency. It defines an Earth-centred, Earth-fixed frame and is the reference used by GPS, globally consistent to within about a metre.
GPS itself began as a US Department of Defense project in 1973. The first prototype satellite flew in 1978 and the full constellation of 24 satellites became operational in 1993. For civilians the signal was deliberately degraded by a policy called Selective Availability until 1 May 2000, when President Bill Clinton signed a directive switching it off. After that an ordinary receiver could fix its position to about 5 m; receivers that use the newer L5 signal can reach about 30 cm. That accuracy is what lets a surveyor with a GNSS receiver on a pole tie a newly mapped building, or a ground control point for an aerial survey, into the global grid.
For a closer look at how a receiver turns satellite timing signals into a position, see our explainer on how GPS works.
Photographing the Surface: Landsat 1 in 1972 to Sentinel-2C in 2024
The base layer of most modern maps is imagery from space. The idea was proposed in 1965 by William Pecora, then director of the US Geological Survey, and the satellite met intense opposition from the Bureau of the Budget, while the Department of Defense feared a civilian programme would compromise the secrecy of its reconnaissance. On 23 July 1972 the Earth Resources Technology Satellite was launched; it was renamed Landsat 1 in 1975. The joint NASA and USGS programme is the longest-running satellite record of Earth, and the most recent addition, Landsat 9, was launched on 27 September 2021. Each Landsat scene covers about 185 by 185 km, and Landsat 7 imaged at 15 to 60 m resolution, returning to the same spot every 16 days.
Europe’s Copernicus programme added the Sentinel-2 satellites, built by an Airbus-led consortium and run by the European Space Agency. Sentinel-2A was launched on 23 June 2015, Sentinel-2B on 7 March 2017 and Sentinel-2C on 5 September 2024. They record 13 spectral bands at 10 to 60 m resolution over all land from 56°S to 84°N, and the data are free. Mapmakers compare images from different dates to spot new roads, felled forest or a changed coastline, then update the vector layers of the map.
Our guide to satellite imagery explains what spectral bands and resolution mean in practice.
Measuring Height: STS-99’s Radar and LiDAR Over the Maya Lowlands
A flat photograph does not show how high the ground is, so elevation comes from separate measurements. In February 2000 the Space Shuttle mission STS-99 spent 11 days flying the Shuttle Radar Topography Mission, a joint project of NASA and the National Geospatial-Intelligence Agency. It carried two radar antennas, one in the payload bay and one on a long mast, so it could measure height by interferometry in a single pass. The resulting elevation model covered the Earth from 56°S to 60°N and was the most complete high-resolution topographic database of the planet until the ASTER model of 2009. It is free to download, which is why so many hillshaded web maps rest on it.
For detail at the scale of a single field or building, mapmakers use LiDAR, which fires laser pulses from an aircraft or drone and times their return. Some pulses slip through gaps in a forest canopy to the ground, so LiDAR can strip away vegetation and reveal the bare surface beneath. In 2016 it traced 17 raised causeways linking the ancient Maya city of El Mirador to other sites in northern Guatemala, and in 2018 a survey of the Maya Biosphere Reserve found more than 60,000 human-made structures, showing that Maya settlement had been far larger than archaeologists had assumed. Earlier, in 2012, the same technique revealed that Angamuco, a Purépecha city in the Mexican state of Michoacán, contained about the same number of buildings as present-day Manhattan.
Roger Tomlinson’s Layers: How a GIS Assembles the Map
The software that combines all these inputs was born in Ottawa. In the 1960s Roger Tomlinson (1933–2014), a geographer working for the aerial survey firm Spartan Air Services, had the idea of joining land-use mapping with the new electronic computers. He went on to plan and direct the Canada Geographic Information System, the first computerised GIS in the world, built to store the data of the Canada Land Inventory, and he is widely called the father of GIS.
The principle has not changed. A GIS stores each kind of information as a separate layer tied to the same coordinates: imagery, elevation, roads, rivers, buildings, place names, property boundaries, census statistics. Layers can be overlaid, queried and analysed together, so the same database can draw a road atlas, plan a flood defence or route a delivery van. In 1969 Jack and Laura Dangermond founded Esri in California with $1,100; by 2015 its ArcGIS products held about 45% of the GIS market. QGIS, a free and open-source alternative, runs on Windows, macOS and Linux.
The final stage is cartographic design. Because there is always more data than can fit, mapmakers apply generalisation, choosing which features to show at each scale and simplifying their shapes. A motorway that appears at every zoom level might be the only road drawn at a continental scale, while footpaths and building outlines appear only when the view is zoomed close in. Our guide to what GIS is goes further into how layers and queries work.
Regent’s Park, December 2004: Steve Coast and the Volunteer Map
Not all modern map data comes from governments or companies. OpenStreetMap was started in 2004 by Steve Coast, then a student at University College London, who was annoyed that the Ordnance Survey refused to make its data freely licensable. The first contribution was a street he entered in December 2004 after cycling around Regent’s Park with a GPS tracker. Volunteers add data from GPS traces, ground surveys and imagery tracing, and large imports have filled gaps, such as the US Census TIGER road dataset completed in October 2007. The data are published under the Open Database License, free for anyone to reuse, and adoption by apps and websites sped up after Google introduced charges for heavy use of Google Maps in 2012.
Growth was fast: 9,000 registered users by July 2007, 100,000 by March 2009 and one million on 6 January 2013, around 30% of whom have added at least one point. The project proved itself after the 2010 Haiti earthquake, when OpenStreetMap and Crisis Commons volunteers traced fresh satellite imagery and charted Port-au-Prince’s streets, buildings and refugee camps within two days. The result was described as the most complete digital map of Haiti’s roads. Relief organisations including the World Bank used it. The Humanitarian OpenStreetMap Team has since repeated the approach for the Northern Mali conflict and Typhoon Haiyan in 2013 and the West African Ebola epidemic in 2014.
May 2007 to 83 Countries: Street View and the Ground-Level Layer
The last layer is the view from the street. Google launched Street View in 2007 in several US cities, using cars with roof-mounted camera masts that photograph a continuous panorama as they drive. By 2017 it covered 83 countries, with some imagery captured on foot, and even by underwater cameras, where cars cannot go. Besides letting users look around, these images give mapmakers details no satellite can see from above: house numbers, shop names, turn restrictions and road signs, which are then written back into the map database.
Put together, the modern method is a loop rather than a single survey: position from GNSS, surface from satellites, height from radar and LiDAR, detail from the street and from volunteers, all fused and generalised in a GIS and republished as soon as something changes. If you want to try the reverse process and read one of those finished products properly, start with our guide on how to read a map.
