The City Before the City

How Urban Digital Twins Could Change the Way We Build

The most useful digital twin may not be the one that shows us the city most beautifully. It may be the one that helps us understand what happens when we change it. There is a point in every digital transformation where better representation stops being enough.

In architecture, we have already crossed that point. We can create highly detailed BIM models. We can build enormous 3D city environments. We can map infrastructure through GIS. We can scan roads and buildings with LiDAR. We can stream information from sensors into dashboards. We can even place all of this inside a real-time 3D environment.

But there is still a fundamental question:

What happens if the city changes?

• What happens if a new township adds thousands of residents?

• What happens if a major road is closed during peak traffic?

• What happens if extreme rainfall exceeds the capacity of a drainage network?

• What happens if an electrical substation fails at the same time that a transport corridor is disrupted?

• What happens if an ambulance needs to cross the city at the worst possible time?

These are not visualization questions.

They are questions about behaviour, dependency and consequences. And that is where urban digital twins become interesting.

A city is not a collection of buildings

When we look at a city from above, it is easy to see it as a collection of physical objects.

• Buildings

• Roads

• Bridges

• Parks

• Railways

• Water bodies

• Utility networks

But the physical objects are only the visible layer. Underneath them is a much more complicated system.

A road carries people. People generate traffic. Traffic affects junctions. Junctions affect emergency response. Buildings consume electricity and water. Buildings generate wastewater and stormwater runoff. Rainfall affects drainage. Drainage affects roads. Road closures redistribute traffic. Power failures can affect buildings, communication systems and traffic signals. The city behaves less like a collection of objects and more like a system of systems. That distinction is important. Because if the systems are connected, studying them separately can sometimes hide the most important part of the problem: what happens between them.

Ahmedabad gave us a real-world example

We do not have to look very far for evidence of how quickly these connections can become visible. On July 23–24, 2026, Ahmedabad experienced an exceptional rainfall event. The city recorded 294.6 mm of rain in 24 hours, making it the fourth-highest 24-hour rainfall total recorded in the city's history. The previous record was 414.8 mm in 1927.

The consequences were not limited to water on roads.

Roads and underpasses were flooded. Residential areas were affected. Transport was disrupted. In Bopal-Ghuma, the aftermath was particularly striking. According to reporting based on Ahmedabad Municipal Corporation information, nearly 30,000 people across more than 30 residential societies were affected. Some areas remained waterlogged for almost three days.

A digital twin could not have stopped the rain. And it would be irresponsible to suggest that it could have predicted every flooded street perfectly. But imagine having a sufficiently detailed digital representation of the area before the event.

It could potentially help answer questions such as:

• Where does water accumulate first?

• Which drainage areas are likely to become constrained?

• Which roads become inaccessible?

• Which entrances to residential communities are vulnerable?

• Which critical facilities are exposed?

• Which emergency routes remain available?

And, perhaps most importantly:

If one road becomes unusable, where does the traffic go next?

That last question takes us beyond flood mapping. It takes us into urban consequence modelling.

A digital twin is more than a 3D model

This distinction is still misunderstood. A 3D city model tells us about geometry.

BIM adds structured information about built assets. GIS adds geographic relationships and spatial context. Simulation models behaviour. Sensors provide observations from the physical environment.

A digital twin brings these pieces into a connected computational environment. There is no single universally accepted definition that makes every digital twin identical, and current research continues to debate architecture, maturity and interoperability. Recent work on systems of digital twins specifically notes that definitions and architectures are still not fully standardised.

So I would avoid thinking of a digital twin as a product. It is better understood as an architecture.

An architecture that connects:

physical assets + digital models + data + simulation + operations + decision-making

The 3D model may be what people see.

But the intelligence is underneath it.

From geometry to behaviour

This is where the role of BIM and GIS becomes particularly interesting. Imagine a building represented in a city model. At the simplest level, it is a 3D object.

With BIM, that object can become much richer.

It can have information about:

• building use

• floor areas

• materials

• structural systems

• mechanical systems

• occupancy

• entrances

• vertical circulation

• equipment

• maintenance information

Now connect the building to GIS.

It has a location. It has relationships with roads, neighbouring buildings, terrain and infrastructure. Connect it to a utility network. Now we know where its water, electricity and communications come from. Connect it to sensors. Now we have observations about how the building is behaving. Connect it to simulation. Now we can ask questions about what might happen under different conditions.

That is a very different proposition from simply having a beautiful 3D model.

The hidden city: infrastructure

Some of the most important parts of a city are the parts we rarely see.

• Stormwater drains.

• Sewerage networks.

• Water supply.

• Electrical grids.

• Substations.

• Transformers.

• Telecommunications.

• Pumping stations.

• Utility corridors.

These systems are difficult to visualise because they are networks rather than isolated objects. A pipe is useful information only when we understand what it connects to. A transformer becomes more meaningful when we know which parts of the electrical network depend on it. A pumping station becomes important when we understand the drainage system around it. This is why the next generation of city models will need to represent relationships, not just objects.

The question is no longer simply:

“Where is the asset?”

It becomes:

“What depends on this asset?”

That is a much more powerful question.

The transportation network is another example

Traffic is often treated as a separate planning discipline. But in reality, transportation touches almost everything. A new development changes trip generation. A road closure changes traffic distribution. A flooded underpass changes route choice. A public transport disruption shifts demand onto roads. A major incident can affect emergency access. A traffic signal failure can turn a functioning junction into a bottleneck. This is why transportation becomes particularly interesting when connected to a digital twin. In 2026, researchers at NYU working with the New York City Fire Department published research on an AI-reinforced Traffic Digital Twin for emergency medical vehicle operations.

Their system combined traffic simulation, real-world data and AI to test emergency-vehicle interventions. In one studied application, optimising ambulance cross-street locations produced up to a 14.5% reduction in expected travel time during morning traffic under the conditions examined.

The important part is not the 14.5%. It is the method.

The researchers were not claiming that digital twins universally reduce emergency response time by 14.5%.

They were demonstrating that a digital twin can become an environment in which interventions can be tested before being implemented in the real world. That is exactly the direction urban planning needs.

Simulation before construction

This may eventually become one of the strongest applications of urban digital twins.

Imagine a developer proposes a large township.

Under a conventional process, the project may require separate assessments for:

• traffic

• drainage

• utilities

• fire access

• environment

• public transport

• infrastructure capacity

  
  

Those studies are important. But what if the proposed development could be introduced into a shared digital representation of the surrounding city?

Then planners could test several scenarios.

Existing condition

How does the area behave today?

Development condition

What changes after the project is introduced?

Peak traffic

What happens during the busiest period?

Extreme rainfall

What happens when the drainage system is stressed?

Road closure

What happens if a major access route becomes unavailable?

Emergency scenario

Can emergency vehicles still reach critical facilities efficiently?

The question changes from:

“Does the development fit?” to:

“What does the development do?”

That is a much more useful question.

India is beginning to move in this direction

There are already signs that this thinking is moving from research into public-sector infrastructure planning

In August 2026, the Government of India approved Phase I of the Urban Flood Risk Management Programme for seven cities: Ahmedabad, Bengaluru, Chennai, Hyderabad, Kolkata, Mumbai and Pune. The total financial outlay is ₹3,075.65 crore.

Phase II covers another 11 cities, with a total financial outlay of ₹2,444.42 crore. These programmes are not digital-twin programmes themselves.

That distinction matters.

But they show something important about the direction of urban infrastructure policy: flood resilience, drainage capacity, road connectivity and urban infrastructure are increasingly being treated as interconnected systems rather than isolated engineering problems. The government says the approved projects are intended to improve stormwater conveyance, reduce flooding and waterlogging, improve road connectivity and increase drainage resilience.

This is precisely the type of environment in which digital-twin technology can become useful.

Not as a replacement for engineering.

As a layer that connects engineering information and helps decision-makers test scenarios.

Chennai is taking a practical first step

Chennai provides an interesting example of how this could happen at city level.

In 2026, the Greater Chennai Corporation proposed a digital-twin pilot covering approximately 5 square kilometres around Nungambakkam, College Road, Greams Road and Anna Salai. The proposed system is intended to combine data from the city's Integrated Command and Control Centre, sensors and GIS mapping, alongside 3D/4D simulation and AI. The wider initiative includes an Urban Data Observatory and infrastructure asset-management systems.

There is another interesting part of the plan.

The corporation has proposed LiDAR surveying of approximately 1,000 km of roads to generate high-resolution spatial information for planning, traffic management and maintenance.

I think the scale of the pilot is actually encouraging.

It is tempting to imagine a city-wide digital twin containing every building, pipe, road, sensor and utility from day one.

That is probably not the most practical way to start.

A better approach may be:

choose a problem → choose an area → build the data foundation → connect the models → test the system → expand.

A digital twin does not have to model everything to be valuable.

It needs to model the right things well enough to answer a useful question.

The technology underneath the visualisation

An urban digital twin is unlikely to be one giant software package.

It will probably be an ecosystem.

At the foundation:

BIM + GIS + survey data + infrastructure databases + remote sensing + IoT

Above that:

common identifiers + semantics + metadata + spatial relationships

Then:

hydraulic + traffic + energy + environmental + structural + evacuation simulations

Then:

AI + optimisation + forecasting + anomaly detection

And finally:

maps + dashboards + 3D environments + scenario interfaces

The difficult part is not creating another 3D viewer.

The difficult part is getting these systems to understand one another.

Why interoperability may matter more than AI

This is one of the less exciting parts of the conversation, but perhaps one of the most important.

A city can have excellent BIM models and still struggle to integrate them with GIS. It can have excellent GIS and still lack reliable infrastructure information. It can have sensors but no common data structure. It can have AI but poor-quality input data.

That is why open standards matter.

IFC 4.3 is now formally published as an ISO standard, with the current official release identified by buildingSMART as IFC 4.3.2.0 / ISO 16739-1:2024. The standard also expands IFC's capabilities into infrastructure domains.

CityGML 3.0 provides a semantic model and exchange framework for 3D urban objects. Its intended applications include urban planning, BIM integration, disaster management, mobility and environmental simulation.

But open standards do not magically create interoperability.

The harder problem is semantic alignment.

Does “road” mean the same thing in two systems?

Does an asset have the same identifier everywhere?

Does a building model know which GIS feature represents the same physical building?

Does a drainage asset in one department connect correctly to the network model in another?

These sound like data-management questions.

In reality, they may determine whether the digital twin works at all.

The research is becoming more realistic

One thing I particularly like about the latest research is that it is becoming less promotional. Researchers are increasingly asking not only what digital twins could do, but whether they actually work at operational scale. A 2026 systematic review of urban digital twins for water-sensitive urban design screened 170 publications and analysed 48 studies in detail.

Its conclusion is revealing: most applications remain at the conceptual or prototype stage, with relatively few operational implementations. Major barriers include data interoperability, model reliability and governance.

Another recent review identifies interoperability and practical value as major challenges in urban digital twins, rather than treating the technology itself as the difficult part.

This is important.

The future of digital twins is not going to be created simply by making bigger models.

It will be created by making models trustworthy enough to support real decisions.

AI has a role — but it is not the whole story

There is a tendency today to put AI into every technology discussion. Digital twins are no exception. AI can be extremely useful.

It can identify anomalies.

It can detect patterns.

It can forecast demand.

It can optimise routes.

It can compare large numbers of scenarios.

It can help operators understand enormous volumes of data.

But AI should not be confused with physical simulation.

If the question is how water will move through a drainage network during a particular rainfall event, physics still matters.

If the question is how traffic will redistribute after a road closure, network structure still matters.

If the question is whether a structure can withstand a particular load, engineering principles still matter.

The strongest urban intelligence systems will therefore probably combine:

real-world observations + physical models + simulation + statistical methods + AI + human judgement.

Not AI instead of everything else.

AI alongside everything else.

A digital twin should not pretend to know the future

This is perhaps the most important principle. A digital twin is not a crystal ball. It is a model. And every model contains assumptions. Rainfall forecasts contain uncertainty. Sensors can fail. BIM information can become outdated. Infrastructure records can be incomplete. People do unexpected things. Construction changes the environment. Traffic demand changes. New buildings appear. Old infrastructure deteriorates.

So a responsible digital twin should not simply say:

“This road will flood.”

It should be closer to:

“Under this rainfall scenario, using the current infrastructure data and stated model assumptions, this area has a high likelihood of flooding.”

That may sound less impressive.

It is actually much more useful.

Because decision-makers need to understand not only the result, but also how much confidence they should place in the result.

The real challenge may be governance

The technology is only one part of the problem.

Who owns the data?

Who updates it?

Who validates it?

Who controls access?

Who maintains the model?

Who pays for the system after the pilot?

Who is responsible when two datasets disagree?

Who approves the assumptions used in a simulation?

Who can use information about critical infrastructure?

These questions are not secondary.

They determine whether the system survives beyond the demonstration stage.

A digital twin that is accurate on the day it launches but becomes outdated two years later is not really a useful operational twin.

It is a snapshot with a 3D interface.

The real challenge is creating a system that can keep learning from the physical city without losing trust in the information it contains.

From designing buildings to testing urban behaviour

This is where I see an interesting future for architecture, BIM and 3D technology.

For a long time, our digital tools have primarily helped us answer:

What should we build?

The next generation may help us answer:

How will what we build behave?

That is a subtle but important change.

An architect may still design a building.

An engineer may still design its structure.

A planner may still design the neighbourhood.

A transport engineer may still model the road network.

A hydraulic engineer may still model drainage.

But their systems could increasingly interact inside a common digital environment.

The result is not the replacement of expertise.

It is connected expertise.

The city becomes something we can test

Imagine a future planning meeting.

Instead of presenting only drawings, the team presents several possible futures.

Option A

The development as currently designed.

Option B

The same development with additional public transport.

Option C

The same development with increased drainage capacity.

Option D

The same development with a different road hierarchy.

Option E

The same development under an extreme rainfall scenario.

The team can compare the consequences.

Traffic.

Flood exposure.

Emergency access.

Energy demand.

Infrastructure requirements.

Pedestrian movement.

Environmental effects.

The question is no longer:

Which design looks best?

It becomes:

Which design performs best under the conditions that matter?

That is a very different way of thinking about urban design.

The future city may be built twice

The phrase “digital twin” can sometimes make the technology sound more complicated than it really is. At its heart, the idea is surprisingly simple. Before making a major physical change, create a sufficiently useful digital representation of the system.

Then test the change.

See what breaks.

See what improves.

See what depends on what.

Change the proposal.

Test it again.

Only then move toward physical implementation.

The objective is not to create a perfect digital copy of a city.

I don't think that is realistic.

The objective is to create a credible computational environment for exploring consequences.

That is a much more achievable—and much more valuable—goal.

The city before the city

We have spent decades improving the way we represent buildings. We are now beginning to improve the way we represent cities. The next step is more ambitious.

We need to improve the way we understand urban behaviour.

That means connecting architecture with BIM.

BIM with GIS.

GIS with infrastructure.

Infrastructure with real-world data.

Data with simulation.

Simulation with AI.

And all of it with the people who actually make decisions.

The most valuable urban digital twin may therefore not be the one with the most detailed buildings. It may be the one that can answer a simple question with enough confidence to influence a real decision:

“If we do this, what happens next?”

That is where I believe the real opportunity lies.

Not in building a digital copy of the city for the sake of having one.

But in creating a place where we can test the city before the city has to experience the consequences itself.

Build it digitally.Test it intelligently.Then build it physically.