Digital twin

An urban digital twin is a maintained digital representation of a city or urban system that stays meaningfully connected to the place it represents. It combines geometry with changing data so people can inspect conditions, test scenarios or support real decisions. A detailed 3D city model is useful, but it is not automatically a twin.

This topic guide connects three layers of the atlas: the plain-language digital twin definition, documented project records, and our evidence-led comparison of Singapore, Tokyo, Seoul and Rotterdam. Together they show why “digital twin” describes an operating model, not a single product.

What turns a city model into a digital twin?

A useful urban twin normally has four ingredients:

1. A defined physical subject. That may be a whole city, a port, a transport network, a district or the infrastructure below a street.
2. A maintained data relationship. Sensor feeds can provide it, but so can scheduled surveys, asset registers and administrative updates. “Real time” is not required for every layer; a documented refresh cycle is.
3. A decision or service. The model should help someone inspect the present, rehearse a change or coordinate work. Photorealism alone is not a use case.
4. Durable ownership. Agencies must know who maintains each layer, who may access it and how the city can move its data if a vendor changes.

This is why the boundary matters. A static 3D rendering is a model. A live dashboard displays observations. A simulation tests assumptions. A digital twin can combine all three, but only when the connection to the physical system is maintained and used.

Four city models worth comparing

The four systems are not competing versions of the same product. Singapore emphasises national coordination; Tokyo makes reuse central; Seoul offers a municipal simulation laboratory; Rotterdam connects the visible city with cables, pipes, foundations and water-related planning. The full city comparison examines scope, openness, coupling, cost and the gaps in public outcome evidence.

A city may need more than one twin

Rotterdam makes that point unusually clear. The municipality's planning model is distinct from the Port of Rotterdam IoT and hydro/meteo platform, which combines sensor observations and prediction models for vessel operations and a future port twin. The Reyeroord Living Lab is different again: a neighbourhood testbed for sensing and public-space governance rather than a citywide twin.

Separating these systems can be a strength. A port authority, planning department and neighbourhood team have different users, security needs, update cycles and measures of success. One universal “city operating system” may create more coordination overhead than value.

What can digital twins help cities do?

Documented uses across the atlas include flood and wind scenarios, solar-potential studies, underground-infrastructure coordination, accessibility analysis, disaster planning, autonomous-vehicle testing and vessel operations. The strongest cases start with a named decision and then choose the model and data needed for it.

The evidence is much weaker on outcomes. Public sources describe coverage, datasets and possible simulations far more often than they report planning time saved, construction clashes avoided, recurring maintenance cost or independently measured damage prevented. That does not make the platforms useless. It means a city should specify outcome measures before procurement and publish what changed outside the model, not only what was added inside it.

Questions to ask before procurement

• Which decision will change? “Improve planning” is too broad; testing a flood route for a named district is concrete.
• What is the update contract? Name the owner, refresh interval and acceptable age for every important layer.
• Who can reuse the data? Internal access, a public viewer and downloadable open data are different commitments.
• Can the city change vendors? Geometry, metadata and scenario logic should be exportable in documented formats.
• What remains closed? Utilities, security-sensitive assets and movement data need explicit access rules.
• How will value be measured? Track use, changed decisions, time saved, stale datasets and recurring cost.

The practical takeaway: start with a decision, not a rendering. The durable asset is the agreement around purpose, ownership, updates, access and measurement. The 3D interface is only one possible view of that system.