The future of cartography is a map that is never finished. Instead of being surveyed once and printed, it is rebuilt continuously from satellite images, airborne lidar, street-level cameras, GPS traces and volunteer edits, with machine learning turning that raw data into roads, buildings and terrain. The results are already visible. Virtual Singapore, a 3D digital twin of the whole city-state, was completed in 2022. Lidar flights over northern Guatemala revealed more than 60,000 hidden Maya structures in 2018. The Seabed 2030 project had raised the share of the ocean floor mapped to modern standards from 6 per cent to 27.3 per cent by June 2025. And the UN General Assembly, voting on 4 September 2026, urged the world to move away from the Mercator projection towards equal-area maps such as Equal Earth. Maps are becoming live, three-dimensional, layered onto what a phone camera sees, and fairer about the sizes of continents.
From Survey Chain to Sensor Stream: How a Map Is Made Now
A modern map forms in four stages: capture, extraction, conflation and publication. Each stage now runs largely on satellites, sensors and software rather than on chains and theodolites.
Capture starts in orbit. The Landsat programme, run jointly by NASA and the US Geological Survey, launched its first satellite on 23 July 1972 and is still the longest-running record of Earth from space; Landsat 9 went up on 27 September 2021. Europe's Sentinel-2 satellites photograph land and coasts at 10 to 60 m resolution; flying half an orbit apart, a pair of them revisits each place every five days, and a third satellite, Sentinel-2C, was launched on 5 September 2024. Elevation came from the Shuttle Radar Topography Mission, an 11-day flight in February 2000 that measured the land surface between 56°S and 60°N using radar interferometry from two antennas, one on a 60 m mast.
Position became precise in the same year. Until 1 May 2000 the civilian GPS signal was deliberately degraded by "Selective Availability", which held ordinary receivers to an average accuracy of about 100 m. When President Bill Clinton had it switched off, accuracy improved to about 5 m. Receivers using the newer L5 band now reach about 30 cm, and survey-grade equipment works to within 2 cm, precise enough for engineering and land surveying.
Extraction is where machine learning now does the work that teams of draughtsmen once did, tracing building footprints and road centrelines from imagery. Filling gaps between measurements still rests on an idea Waldo Tobler published in 1970 in a study of Detroit's growth: "near things are more related than distant things". His first law of geography underpins interpolation methods such as kriging. Conflation then merges sources. The Overture Maps Foundation, launched in December 2022 under the Linux Foundation by Amazon, Meta, Microsoft and TomTom, released its first open dataset in April 2024 and is designed to sit alongside OpenStreetMap rather than replace it.
Under the Petén Canopy: Lidar and the Maya Lowlands of Guatemala
Lidar shows what the next generation of maps can see that photographs cannot. An aircraft fires rapid laser pulses at the ground and times their return; most bounce off leaves, but a fraction slip between them and come back from the soil. Software keeps those ground returns and strips the forest away, leaving a bare-earth model that shows micro-topography such as low walls, terraces and old roads that are invisible from the air.
In 2018 archaeologists flying lidar over the area around the Maya Biosphere Reserve in northern Guatemala mapped more than 60,000 previously uncharted structures: houses, palaces, raised causeways and defensive works. The survey suggested that between 7 and 11 million people lived in the Petén Basin during the Late Classic period, roughly AD 650 to 800, far more than earlier estimates, and showed that cities such as Tikal were larger than anyone had drawn them. An earlier flight in 2016 traced 17 elevated roads radiating from El Mirador.
The same technique found that the Purépecha settlement of Angamuco in Michoacán, Mexico, held roughly as many buildings as modern Manhattan, and in 2024 it revealed the Maya site of Valeriana and a network of settlements in Ecuador's Upano Valley. The cartographic lesson is that forested regions long shown as empty green on maps will be redrawn with roads, fields and towns that were always there.
Virtual Singapore and the Rise of the City Digital Twin
A digital twin is a 3D model of a real place that updates itself with live data. NASA gave the concept its first practical definition in 2010 for spacecraft, and the idea has since spread from engineering to whole cities. Virtual Singapore, launched on 3 December 2014 under the Smart Nation programme and completed in 2022, is described as the first digital twin of an entire country. It was co-led by the National Research Foundation, the Singapore Land Authority and the Government Technology Agency, and built with Dassault Systèmes' 3DEXPERIENCE City platform.
The project grew out of a practical problem. After heavy rain flooded parts of the island in 2011, the government began a 3D mapping programme so planners could see which streets and basements lay lowest. Half the cost came from the government and half from agencies that needed the model, principally the Civil Aviation Authority of Singapore and the Public Utilities Board. A planner can now test how a new tower shades a park, how wind moves between blocks, or where floodwater would pool, before anything is built.
Singapore had predecessors. Helsinki began building an interactive 3D city model, Helsinki Arena 2000, in 1996, and Digital City Kyoto, launched in 1998, fed live camera images from transit stations into its virtual streets. What has changed is scale and refresh rate: the twin is becoming the reference map, and the printed city plan a snapshot taken from it.
Port-au-Prince, January 2010: Volunteers Map a Disaster in Two Days
The future map is also a shared one. OpenStreetMap began in 2004 when Steve Coast, then at University College London, grew frustrated that Britain's Ordnance Survey would not release its data under a free licence. The first street went in during December 2004 after Coast cycled around Regent's Park with a GPS logger; by July 2007 the project had 9,000 registered users.
Its turning point came after the Haiti earthquake of January 2010. Working from available satellite imagery, OpenStreetMap and Crisis Commons volunteers mapped the roads, buildings and refugee camps of Port-au-Prince in two days, producing what became the most complete digital map of Haiti's roads. The World Bank, the European Commission's Joint Research Centre, the UN Office for the Coordination of Humanitarian Affairs and UNOSAT all used it. The Humanitarian OpenStreetMap Team repeated the model for the Northern Mali conflict in January 2013, Typhoon Haiyan in the Philippines in November 2013 and the West African Ebola epidemic from March 2014. When Ebola reached Guinea in 2014, volunteers mapped 100,000 buildings and hundreds of miles of roads there in just five days.
Seabed 2030: Charting the Floors of the Pacific, Atlantic and Indian Oceans
The largest blank on the world map is underwater. The General Bathymetric Chart of the Oceans (GEBCO), run under the International Hydrographic Organization and UNESCO's Intergovernmental Oceanographic Commission, is the only intergovernmental body with a mandate to map the entire ocean floor. In 2016 it launched the Nippon Foundation–GEBCO Seabed 2030 Project to finish the job.
When it started, only about 6 per cent of the seabed had been surveyed to modern standards. By June 2025 the figure was 27.3 per cent. About 14.5 million km² of new depth data entered the GEBCO grid in 2019 alone, and an area the size of Europe was added between 2020 and 2022. The data comes from research ships, navies, companies and even small boats fitted with echo sounders, and the British Antarctic Survey varies its ships' routes so that each voyage crosses unmapped water. The target grid now uses smaller squares in shallow seas and larger ones over the abyss.
Satellites are joining from above. SWOT, a NASA and CNES mission launched on 16 December 2022, is making the first global survey of Earth's surface water, measuring river and lake levels and resolving ocean features 15 to 25 km across, roughly ten times finer than earlier altimeters.
From Street View to Equal Earth: What Everyday Map Users Will Notice
For most people the change arrives on a phone. Google Street View launched on 25 May 2007 in several US cities and covered 83 countries by 2017. In 2019 Google Maps added Live View, which draws walking arrows over the camera image of the street ahead; Immersive View, built from Street View and aerial photographs, launched in February 2023 and extended to routes in 15 cities that October. Lens in Maps, which labels shops and stations seen through the camera, started in Los Angeles, San Francisco, New York, London and Paris in late 2022 and reached 50 more cities in October 2023.
The world map is shifting too. Equal Earth, a 2018 design by the cartographers Bojan Šavrič, Bernhard Jenny and Tom Patterson, shows every country at its true relative size. After the African Union backed the "Correct the Map" campaign, the UN General Assembly adopted a resolution on 4 September 2026, with 164 states in favour, one against and six abstaining, discouraging the Mercator projection for most uses other than navigation. France's foreign minister Jean-Noël Barrot then announced that French schools would stop teaching the Mercator map. The resolution is not binding, but textbooks and software are expected to follow.
The common thread is that the map has turned into a database with many views: a lidar model, a digital twin, a humanitarian layer, an AR overlay. The techniques behind today's maps, the GIS software that stores them and the satellite positioning that places us on them are the foundations, and the long arc from clay tablets to digital twins is traced in our history of cartography.