Mirror worlds: Can digital twins save humanity?

We are taking a look back at stories from Cosmos Magazine in print. This article by Prianka Srinivasan was published in June 2024.

In the physical realm, Tuvalu is under threat.

The Pacific nation, made up of nine atolls dotting a 676-kilometre stretch of ocean midway between Hawai‘i and Australia, is one of the lowest-lying countries in the world – its highest point peaks just a few metres above sea level. Residents fear the waves that constantly lick at the shore will one day swallow their land completely. Some have already been forced to relocate from their coastal homes as droughts, violent storms and floods become more frequent and unpredictable. 

Climate change could soon push their country to oblivion. A recent technical report from NASA reveals Tuvalu is experiencing sea level rise 1.5 times faster than the global average, and predicts that by 2050, much of its land and critical infrastructure will be covered by average high tide levels.

In the digital realm, though, Tuvalu hopes to attain immortality.

Its government plans to replicate the entire nation onto a virtual platform. Te Afualiku – a small islet expected to be one of the first in Tuvalu to be completely submerged – has already been painstakingly mapped, digitised and put on the Metaverse as an interactive simulation by developers from the Australian firm Accenture Song. The team couldn’t visit the islet due to COVID travel restrictions; they constructed the proof-of-concept model “by eye” using drone footage and screenshots sent to them by Tuvalu residents via WhatsApp. It’s hoped that eventually clones of all 124 of Tuvalu’s islands will be accessible online and through virtual-reality headsets.

But the country’s plans extend far beyond simply making three-dimensional copies of their fragile lands. They plan to recreate an entire government on the blockchain, so that all administrative processes, institutional affairs and taxation procedures can happen virtually. Last year, Tuvalu also launched a “Digital Ark” program that will preserve copies of the country’s cultural and historical artifacts on an online database. It’s hoped these projects, collectively called the “Future Now” initiative, will allow Tuvalu’s citizens to operate within a living digital twin of their nation.

“We can use technology to preserve culture, our cultural heritage, our history, our language.”

“Tuvalu is the first digital nation in the sense that we [will be able to] exist fully online without a physical territory,” says Simon Kofe, Tuvalu’s Foreign Minister. “We can use technology to preserve culture, our cultural heritage, our history, our language.”

Minister Kofe and I are speaking over a Zoom video call. We have been trying to organise a time to meet online for weeks, but a giant king tide – the worst Kofe has ever seen – recently flooded the country, cutting electricity to parts of the capital Funafuti. The storm also left newly elected parliamentary members stranded on their home islands, halting the formation of the next government and leaving the country’s leadership in limbo for almost a month, meaning Kofe did not have ministerial authority to speak to me. 

Such events are a reminder of the urgency for Tuvalu to rebuild online, Kofe says. “This gives us a view of what is to come. Things are just going to get worse for us Tuvaluans.”

But the frequent storms and power outages also point to the immense challenges facing the government as it races against time to create this digital twin. Is such an ambitious project even possible, let alone worthwhile?

Enter the mirror world

At first glance, the concept of developing virtual replicas of physical spaces might not seem so groundbreaking. We’ve all used Google Maps or virtual simulators to explore real-world destinations through our screens. 

But digital twins go one step beyond simply being a visual reproduction of our world. They are constantly fed with real-time data – wind speed, weather and traffic information – by sensors in the field, which change the way the virtual image looks and responds. A true digital twin is therefore a synchronous and ever-evolving reflection of its real-world counterpart – a complex universe trapped behind a screen.

NASA says it developed some of the first digital twins in the 1960s, when its space-shuttle simulations were used to plan and execute missions. Other experts in the field say the technology was first proposed at the beginning of the 21st century, when researchers at the University of Michigan suggested a virtual management system to improve manufacturing processes. Since then, the scales of these models have grown impressively, with researchers now creating digital doppelgangers of entire ­buildings, cities and states.

Arguably, the idea of large-scale digital twins was first sparked by Yale computer scientist David Gelernter in his 1992 book Mirror Worlds: or the Day Software Puts the Universe in a Shoebox. In it, he contemplates a future both terrifying and revolutionary, where computers are so power­ful they can “mimic reality’s every move”.

“Such Mirror Worlds promise to be powerful, fascinating, and gigantic in their implications.”

“This is a three-dimensional kind of reflection: The program reaches out and engulfs some chunk of reality,” Gelernter wrote. “Like a child-sized play village modelled precisely on a real town and tracking reality’s every move, the Mirror World supplies a software object to match and track every real one.

“Such models, such Mirror Worlds, promise to be powerful, fascinating, and gigantic in their implications.”

Breakthroughs over the last decades have inched us closer to this future. Supercomputing has given scientists the ability to digest and analyse massive amounts of data, while artificial intelligence and machine-learning systems can ensure the models are extracting the right data to accurately mirror the real world. 

That’s the hope, anyway. The field is still in its infancy – though pulsing with activity. Digital twins are being developed across the world, in almost every industry. Healthcare professionals are looking to create digital twins of human bodies to personalise treatments without cutting the skin. Urban planners are developing virtual cities to improve transport ­systems. And then there are places like Tuvalu, looking to deploy digital twins to better plan for an uncertain future.

One of the most popular uses of digital twins is at this intersection of climate change adaptation and technology. Just as crash-test dummies simulate what happens to a body in a car accident, the hope is for digital twins to accurately predict what will happen to our homes, cities, oceans and countries as our climate systems face radical change.

There has been a swarm of interest in this area – the United States’ National Academies has said digital-twin technology could “revolutionise atmospheric and climate sciences”, while the European Union is creating a virtual replica of the planet to forecast the impacts of a warming climate.

More on that later – first, let’s dive in at the smaller scale.

Don’t wait, simulate

At the University of Pittsburgh in the US, researchers and engineers are testing whether a digital twin of the campus can help them understand how climate change will affect their ­facilities. The work is led by civil engineer Alessandro Fascetti, who says the power of the technology lies in its ability to make predictions on how different climate possibilities may affect the operation of buildings.

“The most sought-after thing right now for this particular application is transitioning to zero-carbon, or at least to lowering carbon emissions, which is the main thing we’re looking at.”

His team have begun by digitally replicating one building, the Mascaro Centre for Sustainable Innovation, chosen for the vast number of sensors that already mark its walls, constantly collecting data on energy use, occupancy, temperature levels and other variables. 

The researchers have also been busy building the virtual platform to house this data. Fascetti and his research students use mobile lasers – black glass cloches about the size of a small flowerpot – to take images and corresponding spatial information about the building.

“The scanner houses an array of sensors, from 360° cameras similar to the ones on Google Maps cars, to infrared imaging,” Fascetti says, then points to the black lens at the centre of the dome in his hand. “At the same time, this object here in the middle is a LiDAR sensor that collects high-resolution data.”

From here, the possibilities are endless.

LiDAR, or Light Detection and Ranging, is a way of collecting geospatial data by shooting pulses of light out from a laser to an object. 

“You read the time it takes for the reflection to come back, and since you know that the laser travels at the speed of light, you know the distance.”

With this information, Fascetti and the team craft a “digital shadow” by importing the data into a graphics editor called Unreal Engine – the same software used by video game developers. The software converts these millions of data points into an interactive, high-resolution visual model of the building.

“This shadow gets morphed into a twin when we start including all the data streams and predictive models,” Fascetti explains. For example, when temperatures rise in the physical building, this change will simultaneously be represented on the 3D model.

Just like that, a digital twin is born. 

But then comes the hardest part– getting the twin to make accurate predictions about the future. This is done through an “alignment” process, Fascetti says – the researchers will intentionally hide certain data streams and get the model to guess the missing information over many iterations. Once it does so correctly, they’ll know it is capable of making accurate predictions about the physical world.

From here, the possibilities are endless. They can start inputting climate projections and see how the building’s twin reacts. How much extra electricity is necessary to keep classrooms cool when temperatures rise? How will building occupancies change as weather patterns begin to shift?

“We don’t have to wait and see. We can simulate,” Fascetti says.

Even at this small scale, there is something almost mystical in what these engineers are trying to create: a system that will allow us to peek into our possible futures. Until recently, even contemplating such technologies was difficult – the sheer amount of data and computing infrastructure needed simply didn’t exist. Even now, despite his optimism, Fascetti is aware of the challenges.

“If you’re talking even of a medium-sized city, this becomes daunting,” he says. “If you talk about the region – well, at this point, we really don’t know if we even have computers to do that.” 

There are scientists, though, who are trying to find out.

Cloning our cities

In 2015, an aircraft flew above Singapore with a very unusual passenger on board. Operated by the geospatial service AAM Group (now Woolpert), the plane carried a sophisticated LiDAR imaging system that bounced laser beams across the country.

The aircraft was commissioned by the country’s land services department as part of its plans to create a digital twin of the entire nation. The aerial images would be combined with data collected by laser-equipped cars that traced every street in Singapore. Three million panoramic images and around 25 terabytes of data went into the system.

The SGD$73 million National 3D Mapping Programme was conceived to help the country better respond to emerging climate threats, like the flash floods that regularly wash through city streets after heavy rains. Singapore is one of the most densely populated countries in the world, so figuring out how to build infrastructure to best assist its citizens can challenge city planners. It was hoped a digital twin could help take the guesswork out of social, economic and environmental intervention.

“The software offers visualisations of 3G/4G network coverage areas; simulations of crowd control and evacuation measures; and planning scenarios for delivering municipal services, ­analysing pedestrian flows, as well as projecting science research outcomes,” Singapore’s Land Authority said at the project’s launch a decade ago.

Touted as the world’s first digital twin of a country, the 3D simulation of Singapore is exquisite in its detail. Any point in the country can be inspected in 360° of clarity. Users can fly over the city model like a virtual drone. 

The model has allowed city planners to identify flood-prone areas and create a tailored coastal protection plan. Singapore’s 3D building models have also been used to establish a national “solar potential map” that reveals suitable rooftops for solar panel installations. Even the ­country’s parks department is using the digital twin to identify which trees are obstructing motorists and need pruning.

“We are constantly looking at how we can harness the potential of geospatial data and technologies further to support Singapore’s sustainability efforts,” Singapore’s Land Services department said in an email, calling the future of digital-twin technology “limitless”.

“As the scale of digital twins increases, engineers have a difficult balancing act to maintain.”

Digital twins are also in the works for Dubai, Wellington, London, Paris, Melbourne and ­dozens of other places. Experts, like infra­structure engineer Abbas Rajabifard from the University of Melbourne, say that digital twins offer decision-makers the seductive ability to witness the impacts of climate change virtually, before they confront them in reality.

“If we bring this [digital twin] system to life, it becomes like a live testbed – you can bring anything into it, and it provides the solution,” Rajabifard says. He gives the example of planning your morning commute. The simulation would not only tell you if it will rain today, but also the impact of driving versus taking the train – how much time you might save, what the road conditions will be like, how much fuel your trip will consume.

“You can put yourself into that situation virtually … and then you can choose your option,” he says.

But there is some danger behind this hype. As the amount and complexity of information fed into the digital twin grows, and its engineers rely on artificial intelligence models to extract useful information, it will become harder to understand how and why the twin makes its predictions. There’s a risk a digital twin could be treated more as an impenetrable digital oracle.

Rajabifard and his colleagues call this problem the “black box” of digital twin and AI development. “In some areas, [a prediction] can be totally meaningless until the system becomes more mature,” Rajabifard says.

For example, a predictive, AI-powered digital twin used in farming may prioritise a larger harvest over worker safety, without the end-­users knowing what it’s doing. Rajabifard says governments must ask themselves an important question.

“How can we validate that information before we apply it to our decision processes?”

The answer lies in developing powerful “auditing” systems, Rajabifard says – though there’s still “more room to learn” about what those systems might look like. Most likely, it would mean widening the type of data the model wrestles with – in the above example, an auditing system could ensure variables around employee wellbeing, like rates of injury or working hours, are provided alongside the twin’s farming recommendations.

But the solutions are not all technology-­related. Rajabifard has been developing workshops for community and government leaders on how to use digital twins, comprehend their outputs and validate their simulations.

“Let’s engage as much as we can with different authorities so that they can bring their own data sets into this,” he says. 

Planet No. 2

There is a further problem presented by the vast sea of data needed by digital twins to make accurate predictions.

Take the butterfly effect: chaos ­theory’s thought experiment first ­proposed by mathematician and meteorologist Edward Norton Lorenz. It holds that miniscule changes to our weather systems can have massive yet unpredictable flow-on effects – like how the flapping of butterfly wings can eventually lead to a tornado on the other side of the world. 

As the scale of digital twins increases, engineers have a difficult balancing act to maintain. At high resolutions, the twin is able to better take into account granular interactions at the level of butterfly wings, but this would require an explosion in data and computational costs. On the other hand, a lower-resolution model would require less data and therefore cost less – but the twin would also quickly lose its synchronicity with its physical doppelganger, and its powerful prediction capabilities would be greatly diminished.

Peter Dueben from the European Centre for Medium Range Weather Forecast is part of a new initiative to create a digital twin of the entire planet. He is very familiar with the complications posed by the butterfly effect.

“That’s one of the reasons why it’s getting more and more complicated to make good predictions as we go into the future,” Dueben says. “The degrees of freedom are overwhelming.”  

The European Commission-funded project, called Destination Earth – DestinE for short – is combatting this problem by using highly sophisticated sensors and massive computing power to wrangle the “overwhelming” amount of information needed to create a virtual planet.

The MareNostrum 5 supercomputer, unveiled in Barcelona last year and ­capable of executing 314 million billion calculations per second, will be tasked with analysing the data needed to create our planetary twin. Dueben says simulations with the highest resolution will include more than 250 million horizontal grid points, 137 vertical levels and at least 10 different prognostic variables per grid point – which include things like temperature and pressure. 

“It’s something that a normal human can’t really comprehend,” he says.

“What would happen if the rainforest in the Amazon was to disappear? You can look at how it would work.”

But if the team pulls it off, DestinE could supercharge our ability to visualise our climate futures. Current forecasts run at the nine-­kilometre range, at best covering large suburbs or townships, predicting the weather over the next week or so. Meanwhile our existing climate models analyse components like atmosphere chemistry, oceans, land surface and ice to provide broad, global temperature predictions years or decades into the future.

DestinE would provide much greater detail over larger timeframes and smaller areas. Its scientists are aiming to push enough data into the system – from satellites, weather stations and sensors around the world – to develop a model with a powerful one-kilometre grid resolution of our meteorological system. At these higher resolutions scientists would be able to pinpoint paths storm clouds might take as they form over ­villages in the Pacific, or determine risk levels of bushfires in Australia before they even strike.

“If you go to the one-kilometre range of resolution, you basically end up with a model simulation of the atmosphere that is very hard to distinguish from the observation,” Dueben explains; if you were to take a satellite and ask it to focus on a one-kilometre-square patch of land, the images it produces would be identical to what the digital twin simulates.

A digital twin of Earth can help us understand our planet’s past, present and future – but to create such an in-depth replica requires a multitude of smaller twins of the Earth’s systems; for example, physical systems like the reconstruction of Antarctica’s hydrology. Data-fed models of forests, oceans, river systems and more will be crucial to creating a responsive, whole-Earth digital twin. Credit: Earthwave

Two years into the project, Earth’s digital twin is still early in its lifecycle. It’s still unknown precisely how the system will be used, and by whom.  But Dueben believes, ultimately, DestinE can empower governments and policy makers around the world to prepare for climate-changed futures.

“What would, for example, happen if the rainforest in the Amazon was to disappear?” Dueben asks. “You can … look at what the Earth would actually respond to and how it would work.”

This visual component to the digital twin can’t be overstated. It’s true, complex climate modelling is already available to us, including studies into how deforestation can change our communities and the world. But Dueben explains that DestinE, and ­digital twins like it, could allow anyone  to witness these impacts with their own eyes.

“It’s not only about the model development, but also about how we make the data available to users and how the society can interact with the model simulations as well,” Dueben says.

The next phase of the project is to embed powerful machine-learning technologies into the simulation.

‘Worst-case scenario’

While digital twins offer some countries a revealing glimpse into their future (and with it, the possibility to alter its course), for small island countries, those dire ­predictions are already coming true.

In Tuvalu, leaders don’t need technology to witness the impacts of climate change – they can just look out the window. “Certain areas that used to be land are now underwater. We’re also seeing salt water seeping through the land, which is making it very difficult for us to grow things on the island,” Foreign Minister Kofe says.

I ask Dueben if the money and attention put into cloning the planet is really worth the cost, given that the science is conclusive around the impacts of increased fossil fuel emissions. Extreme weather fluctuations, major biodiversity loss and food insecurity have already been predicted by the Intergovernmental Panel on Climate Change, without the need of a digital twin.

“We know basically that it’s going to be bad if we increase climate change. But we don’t know exactly what’s going to happen in our local area,” Dueben responds. He says providing such ­localised images of the future can also be an important tool for communities and governments to understand the impact of climate change, and advocate for a better response.

“Scientists are predicting that our islands could be fully submerged within a matter of decades. This is a plan for that worst-case scenario.”

But what happens if our environments are already facing extinction, or if we are accelerating too fast down a path of climate collapse?  Such questions are front of mind for Kofe. According to him, Tuvalu’s government is using digital-twin technology to preserve an image of the nation today, rather than imagine possible disasters tomorrow.

“Part of our advocacy and messaging is to try and get people to understand how climate change is really affecting countries like Tuvalu that are at the forefront,” he says.

There’s frustration in Kofe’s voice when I speak to him about how technology can help his country. “The media likes to put the attention on the Metaverse stuff but the core of it is just looking at how we can harness the power of technology to improve the lives of Tuvaluans,” he says.

I ask Kofe if he believes developing a digital twin is really a viable solution to the country’s climate change vulnerabilities. Does he really expect Tuvaluans to relocate to an online, virtual country and abandon their physical homes?

On one hand, he hopes contemplating such a future serves as a wake-up-call to the rest of the world, allowing them to avoid entering the digital twin altogether.

“We feel that the more people understand the situation that we’re facing … hopefully that will have a chain reaction to the leadership in their countries,” Kofe says. “Pressure can be put on the leaders to take stronger climate action.”

But he also says his government’s digital twin endeavours aren’t simply “PR stunts”. The country is legitimately preparing for what could happen when their land disappears.

“Scientists are predicting that our islands could be fully submerged within a matter of decades,” Kofe says. “This is a plan for that worst-case scenario.”

Kofe doesn’t know when the government will finish creating Tuvalu’s digital twin. The plans are, after all, ambitious – to preserve an entire country’s history, culture and geography virtually. Is it fair to ask a vulnerable nation to ­consider such a future for its people? Can a digital twin provide more than a shadow of its reality?

Such questions can only be answered as twins become better at mimicking our real worlds. Engineers in Europe expect the “full digital replica” of Earth to be completed in 2030. Climate scientists predict that around 2030, global temperatures will exceed 1.5°C above pre-industrial levels, pushing many countries into irreversible peril.

At that stage, we may all be faced with the same “worst-case scenario” that Tuvalu contemplates today: pondering if a mirror world can ever truly replicate our real one once it becomes uninhabitable.