As the new U.N. report on climate change makes clear, the perils are growing worse, with storms, floods, and extreme weather events displacing millions; droughts killing crops and exacerbating water scarcity; and increasingly frequent wildfires ravaging the planet.
The report concludes that nations aren’t doing enough. One of the greatest challenges is understanding what’s in store if we don’t take more aggressive, smart, and strategic action. But even if leaders unanimously agreed, where and how should they start?
Such challenges underscore the need for a comprehensive understanding of climate science. Without thoroughly understanding the planet’s interrelated dynamics, it’s difficult to predict the impact of climate change accurately, and which mitigation measures should be prioritized.
The Earth needs a time machine. Fortunately, we now have the technologies to build one, using artificial intelligence, accelerated computing, and science. A digital twin of the Earth will be our time machine.
Can you say zettaflop?
Scientists at the National Center for Atmospheric Research (NCAR) have spent decades developing climate models. To accurately model the Earth’s climate – including the behavior of clouds, storms, air and ocean currents, forests, grasslands, wildfires and ice sheets – they need a supercomputer 1000x more powerful than what they have.
That would require a zettascale supercomputer capable of a sextillion calculations per second (1021) or waiting decades for Moore’s law to yield chips that are 1,000x faster than today. For the challenges we face right now, neither option will work.
But an AI-enabled digital twin may help in the interim, and there are several initiatives already underway.
A scientific consortium in Europe is working on a high-fidelity digital twin called Destination Earth (DestinE) that will model and simulate weather and climate patterns. It will focus on climate change, water and marine environments, extreme weather events, and possible mitigation strategies. The group believes it will help predict climatic trends and disasters with “unprecedented fidelity and reliability.”
NVIDIA also announced plans to build a digital twin of the Earth, called Earth-2.
The digital twin system would combine the company’s Omniverse platform with NVIDIA Modulus, an AI framework that is informed by the laws of physics and learns from streaming data to predict the impact of climate change across an entire region. Powered by an AI-supercomputer of unprecedented scale, this system will enable the community to make a giant leap and accurately model the Earth. The company is expected to provide an update on the digital twin Earth system and climate science technology at its upcoming GTC conference in March.
If a digital twin can be built that would model the enormous, complex dynamics of the entire planet, imagine the number of applications that might benefit from this technology.
Digital twins in action
Thanks to the proliferation of cameras, sensors, and graphic processors, digital twins are being deployed across multiple industries. They are being used in medicine to study internal organs; by industrial designers to create car and aircraft prototypes including Air Force fighter jets; and by architects to design buildings that fully model interior spaces and their environmental systems.
One early adopter, General Electric, has more than 1.2 million digital twins of physical assets such as jet engines, wind farms, offshore oil rigs, pumps and power generators. The company says that its real-time monitoring capabilities have saved customers more than $1.5 billion in reduced downtime and maintenance costs.
BMW Group is using digital twins to accurately recreate entire factories. Since nearly all of the 2.5 million vehicles that BMW produces each year have custom requirements, the company is constantly looking for ways to optimize production.
Bentley Systems, whose software is used by architects to design bridges, utilities, and large industrial buildings, uses digital twins to enable customers to virtually explore massive industrial plants and offshore facilities as if they were walking through physical structures in real time.
To learn more, register for NVIDIA’s upcoming GTC conference, which will offer 15 sessions about digital twin technologies. Representatives from Bentley Systems, BMW, Epic Games, Microsoft, NVIDIA, Siemens Energy, the U.S. Navy and Walt Disney Studios will describe a variety of digital twin applications and explain how they were developed.
The free, virtual conference, which takes place March 21-24, will feature more than 900 sessions on a variety of computing and AI technologies.