How digital twins of Earth are spinning up safer decisions for the planet
Typhoon Chanthu lashed East Asia in 2021. At one point, winds reached 170 mph, making it one of the strongest storms of the season. It wreaked havoc on the region, triggering evacuations and disrupting travel.
During a storm like Chanthu, meteorologists and government leaders want to figure out what's coming next. The future of improving such predictions could lie with the technological approach a company might use to anticipate a hiccup in its supply chain: digital twins.
An increasing number of companies and organizations are spending time and money to create digital twins of the Earth. Driven in part by advancements and improvements in technology like computing power and storage -- as well as the desire to gain a deeper understanding of data -- insight from digital twins could help guide decisions about extreme weather response, urban planning, infrastructure, insurance, and much more.
But much like the physical Earth, its digital twin comes with its share of complexity.
Mirroring Earth before disaster strikes
The reason anyone might want to build a digital twin of the Earth isn't that different than why a company might build a twin of one of its factories or warehouses.
"It's really about business process, or business process transformation using technology," said Alfonso Velosa, analyst at IT consultancy Gartner.
Whether it's a research institution or a business, building a twin of Earth should ideally have a specific purpose or answer a specific question.
Also: How you can use Google Maps to track wildfires and air quality
"We want to be making investments, whether those are on a policy scale or on an infrastructure investment scale, that are really optimized," said Christine Angelini, an associate professor at the University of Florida's environmental engineering sciences department.
Angelini, along with a much larger multi-disciplinary team that includes Karla Saldaña Ochoa, director of the Spatial Human-Centered Artificial Intelligence Research and Experimentation Lab, are working on a project to create a digital twin of Florida. In these early days of the initiative, which received $1.75 million in strategic funding from the Florida Legislature in 2023, they're digitizing the city of Jacksonville and focusing on the issues of housing hazards and health.
Similarly, Michael Seablom, head of NASA's Earth Science and Technology Office (ESTO), talked about his team's work using digital twins to mirror human-made systems, like power grids or roadways. They want to examine how something like climate change could affect those human systems over time. The underpinning questions ultimately involve Earth science.
Also: Artificial intelligence, real anxiety: Why we can't stop worrying and love AI
For example, ESTO could create a twin to model the water cycle in an area and show how it's affected by other Earth systems. A project like that could yield insight into how events like floods and droughts impact infrastructure, as well as people's lives and property.
In 2022, NASA announced it would support 28 proposals along these lines for two years, for a total of $31 million.
Playing out scenarios that haven't happened, or could possibly happen, is one of the key use cases for a digital twin.
For researchers, it could answer questions about how wildfires spread and where they'll burn next, how typhoons or hurricanes might make landfall, or where to place infrastructure.
"If you have a road that you want to build in a coastal area, what is that coastal area going to look like in 20 years?" said Seablom, who is also associate director of technology for NASA's Earth Science Division.
When digital twins make a world of good
To be sure, digital twins are not new. The lore goes that NASA first conceptualized digital twinning in the 1960s, even if the term itself didn't come into being until decades later. While experts may quibble on in-and-out bounds of what exactly a digital twin is, the basic idea is that it's a virtual mirror of an asset, process, person, or the like that's updated with data that reflects the physical thing. It's not merely an unchanging representation.
In the past few decades, companies and organizations have created digital twins of factories, warehouses, products, supply chains, and more. The Charles de Gaulle airport in France uses sensors and LiDAR to anonymously track travelers through the airport to address snags and clogs in the passenger experience.
The gaming industry proved to be an inadvertent booster of digital twins. The early 2000s saw a boom in what's called intelligent environments, mostly applied to game engines, said the University of Florida's Saldaña Ochoa.
"To create an intelligent environment was to create an environment that is reactive to a user player," she said. "This has been pushed forward from the game engine companies or from the game developers that made available some software that now is useful for other fields of study."
Also: AI agents are the 'next frontier' and will change our working lives forever
Another contribution from the gaming industry came in the form of more powerful graphics processing units, or GPUs.
"They really changed the nature of gaming and professional simulation and just visualization across the board," said Rick Brooks, global director of simulation for BlackShark.ai, a company that uses real-time satellite imagery, combined with a generative AI engine it built called Orca Huntr, to create a photo-realistic digital twin of the Earth (Synth3D) that's been used in a variety of capacities, including Microsoft's most recent version of Flight Simulator.
Blackshark.ai's Orca Huntr technology can create buildings based on what they typically look like in a specific area of the world, including the ones in Paris shown here.
Thanks to the dawn of the Internet of Things (IoT), there's also more data now than ever before -- even if that data isn't always in the best shape or easily accessible.
A presentation from ESTO about the Earth digital twin initiative from 2023 pointed out that NASA's been collecting data on Earth for more than 50 years -- data from space, airplanes, balloons, and sensors. And there's more data to come. Thanks to previously mentioned tech advancements, as well as artificial intelligence and machine learning, it's faster and more possible to extract insights from all the data.
Digital twins lifting off with generative AI
Generative AI is also making its way into the world of Earth digital twins.
For one, it can play a role in enhancing the visual representation of the Earth -- not every Earth digital twin has a visual component.
BlackShark.ai's Orca Huntr generative AI engine handles the task of representing the roughly 1.5 billion buildings on the face of the planet, among other features. What someone might see isn't a brick-for-brick replica of every building, but rather a representation based on information from satellite imagery. The engine looks at roof lines, types, and materials, and applies the characteristics of buildings typical to the area. It can also use the lengths of shadows in the imagery to determine the height of buildings.
This aerial shot of Manhattan shows just some of the 1.5 billion buildings generated with Orca Huntr.
Blackshark.ai has since spun out Orca Huntr as its own product. The company is also using its Orca Huntr tech in the effort to combat wildfires. Blackshark.ai was selected to participate in the XPRIZE Wildfire Competition, which aims to spur tech development to innovate on how wildfires are detected, tracked, and managed -- before they become highly destructive.
Also: Nvidia will train 100,000 California residents on AI in first-of-its-kind partnership
From Nvidia comes a generative AI model called CorrDiff, introduced at the company's GTC conference in 2024. CorrDiff uses generative learning to help predict -- faster and more efficiently -- the finer-scale details of a weather event that might be missing from more traditional models.
Bhoomi Gadhia, senior product marketing manager, said Nvidia worked to build trust in the accuracy of the generative AI predictions by checking the predictions against national prediction models and against what actually happened after the fact to show they were correct.
And in August, Nvidia introduced StormCast, a new generative AI model the company said can make reliable predictions at something called mesoscale, which refers to scale bigger than storms, but smaller than cyclones.
To infinite Earths and beyond?
While the efficacy of creating digital twins of the Earth largely remains to be seen, researchers hope that the impact of these digital twins will be far-reaching.
Aside from the noble aspirations of digital twins, which could help make crucial decisions about how to battle wildfires, evacuate endangered communities, and fight climate change, Gartner's Velosa noted there's product potential here, too. A company, for example, could one day sell a city its twin or pitch the twin to insurance companies.
The market for simulation digital twin-enabling software and services is projected to hit $379 billion in the next decade, up from $35 billion in 2024, according to a Gartner report. There's plenty of room for the industry to evolve. Gartner noted that there are no dominant players yet because of the wide variation in combinations of products and solutions required to make a digital twin for any particular use.
Nvidia is working on three APIs -- training as a service, visualization as a service, and simulation as a service -- as the next milestones for Earth-2. Gadhia said they have use cases from figuring out flight paths to tracking flower delivery on Valentine's Day.
Also: AI can help businesses design for positive externalities
And NASA, Nvidia, Blackshark.ai aren't the only organizations working on building twins of the Earth. These efforts can come as collaborations between research organizations, universities, governments, and companies. For example, the European Union is funding an initiative to create a digital twin of the Earth called Destination Earth, or DestinE, for short. In partnership with the European Space Agency, the European Commission, and many others, the twin launched in June, with the goal of addressing everything from air quality issues to tracking global fish migration and habitats to pollution in polar regions.
Still, despite the projected growth and interest in the possibilities ahead, creating reliable digital twins presents challenges.
Angelini said one of the big goals is to be able to aggregate different types of data and look for interconnections. However, she noted that there can be a huge disparity in the quality and reliability of data from different sources. Secondly, getting data in literal real-time will require further evolution in technology, she said. Both Angelini and Saldaña Ochoa talked about the need for better data aggregation methods.
NASA's Seablom also pointed out that the twins they're working on are not full, exact replicas -- a system like the Earth's atmosphere is simply too complex.
Additionally, finding a way to effectively visualize the twins can be another hurdle.
For Seablom's team, they hope to create a common architecture to unite the different models in the future.
"It's simple to say -- a digital twin of the Earth. But it's a many-years-long project," Gadhia said.