Using the omniverse to understand climate change

Cambrian AI analyst Alberto Romero contributed to this article.

One of the greatest challenges facing humanity today is climate change. Although climate changes occur naturally, human activities over the past 200 years have directly affected otherwise normal climate cycles in a very precise way: We’re heating the world. Scientists cannot warn policymakers enough about the catastrophe we will face in the coming decades unless we act effectively and with the utmost urgency.

NVIDIA CEO Jensen Huang alluded to this issue in his 2022 GTC keynote. He mentioned that scientists believe we need billions of dollars in computing power to “effectively simulate regional climate change.” NVIDIA has developed three core technologies – GPU-accelerated computation, physics-informed AI models, and AI supercomputers – that together can enable millions of times faster. Now the company embarks on a journey to address a crucial goal for climate science: To close the gap to that billion X, we must adopt effective mitigation and adaptation strategies and save our planet from its imminent future.

NVIDIA’s full-stack approach is key: GPU-accelerated, physically-informed, data center-scale AI models are our best shot at meeting the EU mandate to achieve carbon neutrality by 2050. How do you intend to achieve this? With a digital twin of our planet. You will build the most powerful AI supercomputer developed for climate science: Earth-2. It will be a physically accurate, high-fidelity, ultra-high-resolution replica of the Earth that will run continuously to predict climate and weather events on a regional and global scale.

In a Q&A with analysts, Jensen Huang said Earth-2 will be “constantly recalibrating against Earth…constantly refining its models…constantly making future predictions.” And that is exactly what makes Earth’s digital twin uniquely suited to take on the challenge. In contrast to conventional simulations, a digital twin is constantly synchronized with its real counterpart through detailed measurements. It’s also extremely fast. As David Matthew Hall, a senior data scientist at NVIDIA, puts it, Earth-2 can provide “actionable feedback in actionable time.”

“Economists, biologists, scientists and companies in countries around the world will be able to … simulate whether the dams they build in Venice will make a difference. Simulate… what will happen to the Mekong in Southeast Asia. Simulate … what will change in California climate conditions,” Huang said. “Earth-2 will literally be Earth-two.”

There are two main technological pillars to build Earth-2: Modulus and Omniverse, NVIDIA’s simulation engine. The virtual worlds that humans build in the omniverse obey the laws of physics, can be operated at scale, and are shareable—allowing for simultaneous collaboration among scientists around the world. Omniverse enables visualization and interactive exploration, but scientists also need an AI model to simulate and predict Earth’s climate and weather.

That’s what Modulus is for. It is a framework for developing physics ML neural network models trained on terabytes of data. Such a model can then become a replacement for the digital twins. Modulus uses the neural Fourier operator framework to give AI models the ability to understand physics and lay the foundation for predicting climate variability and atmospheric events quickly and accurately.

The simulation and visualization power of the Omniverse, in partnership with the accuracy of Modulus-developed physically-informed AI models, together with data center-scale GPU-accelerated computation, puts NVIDIA in an exceptionally good position to solve the problem of forecasting normal and extreme weather and climate events in ultra-high resolution – at both regional and global scales.

As a first step towards Earth-2, NVIDIA has introduced a digital twin for weather forecasts: FourCastNet. It is a GPU-accelerated physics ML model trained on 10TB of ground data to predict problematic atmospheric events such as hurricanes and atmospheric fluxes. It is the first deep learning model to outperform most modern numerical models (especially the ECMWF’s integrated forecasting system) in precision and speed – up to 5 orders of magnitude faster. In Huang’s words, “What a classic numerical simulation takes a year now takes minutes.”

looking ahead

This is just the beginning for Earth-2. NVIDIA is dedicated to helping the scientific community—and by extension, all of humanity—solve one of the greatest and potentially most existential challenges of our time. And just so you can put in perspective why so much effort is required to solve climate change – and why we shouldn’t treat it like another problem – I’ll conclude by repeating the words Jensen Huang wrote about Earth-2 in 2021 has : “All the technologies we have invented up to this moment are needed to make Earth-2 possible. I can’t think of a bigger or more important benefit.”

disclosure: This article represents the author’s opinion and should not be construed as a recommendation to buy or invest in the companies mentioned. Our firm, Cambrian AI Research, is fortunate to have many, if not most, semiconductor companies among our clients including Blaize, Cerebras, Esperanto, Graphcore, IBM, Intel, NVIDIA, Qualcomm Technologies, Synopsys and Tenstorrent. We have no investment positions in any of the companies mentioned in this article, nor do we plan to open one in the near future. Visit our website at for more information.

Leave a Reply

Your email address will not be published.