New NVIDIA Earth-2 Generative AI Foundation Model Simulates Global Climate at Kilometer-Scale Resolution

With a more detailed simulation of the Earth’s climate, scientists and researchers can better predict and mitigate the effects of climate change.

NVIDIA’s bringing more clarity to this work with cBottle — short for Climate in a Bottle — the world’s first generative AI foundation model designed to simulate global climate at kilometer resolution.

Part of the NVIDIA Earth-2 platform, the model can generate realistic atmospheric states that can be conditioned on inputs like the time of day, day of the year and sea surface temperatures. This offers a new way to understand and anticipate Earth’s most complex natural systems.

The Earth-2 platform features a software stack and tools that combine the power of AI, GPU acceleration, physical simulations and computer graphics. This helps enable the creation of interactive digital twins for simulating and visualizing weather, as well as delivering climate predictions at planetary scale. With cBottle, these predictions can be made thousands of times faster and with more energy efficiency than traditional numerical models, without compromising accuracy.

Leading scientific research institutions — including the Max-Planck-Institute for Meteorology (MPI-M) and Allen Institute for AI (Ai2) — are exploring cBottle to compress, distill and turn Earth observation data and ultra-high-resolution climate simulations into a queryable and interactive generative AI system.

cBottle was field-tested at the World Climate Research Programme Global KM-Scale Hackathon. The event was organized across eight countries and 10 climate simulation centers with the goal of advancing the analysis and development of high-resolution Earth-system models and broadening access to high-resolution high-fidelity climate data.

Revolutionizing Climate Modeling With AI

Climate informatics is traditionally time-, labor- and compute-intensive, requiring sophisticated analysis of tens of petabytes of data stores.

cBottle, incorporating NVIDIA GPU acceleration and the highly optimized NVIDIA Earth-2 stack, uses advanced AI to compress massive amounts of climate simulation data. It’s capable of reducing petabytes of data by up to 3,000x for an individual weather sample — translating to a 3,000,000x data size reduction for a collection of 1,000 samples.

cBottle was trained on high-resolution physical climate simulations, as well as measurement-constrained estimates of observed atmospheric states over the past 50 years.

The model can fill in missing or corrupted climate data, correct biased climate models, super-resolve low-resolution climate data and synthesize information based on patterns and previous observations. cBottle’s extreme data efficiency enables training on just four weeks of kilometer-scale climate simulations.

Global Collaboration for Planetary-Scale Impact

Leading climate institutions are using NVIDIA Earth-2 to advance climate simulation.

MPI-M has tapped Earth-2 to pioneer kilometer-scale climate modeling, using its ICON Earth system model. Harnessing NVIDIA GPU acceleration and performance optimizations, MPI-M researchers led a team that performed the first ever kilometer-scale simulations of the full Earth system, simulating and visualizing Earth’s climate with remarkable detail.

“In the face of a rapidly changing climate, the latest progress with Earth-2 represents a transformative leap in our ability to understand, predict and adapt to the world around us,” said Bjorn Stevens, director of the Max Planck Institute for Meteorology. “By harnessing NVIDIA’s advanced AI and accelerated computing, we’re building a digital twin of the planet — marking a new era where climate science becomes accessible and actionable for all, enabling informed decisions that safeguard our collective future.”

Ai2 and NVIDIA are collaborating to accelerate and enhance climate modeling using the Earth-2 AI stack and GPUs, focusing on making climate simulations faster, more energy efficient and more accessible at high resolutions. This is critical for scientific research and practical applications in weather prediction and climate resilience.

“Planning for climate change challenges societies worldwide,” said Christopher Bretherton, senior director of climate modeling at Ai2. “cBottle is an elegant use of generative AI and an exciting new resource for efficiently simulating local extreme weather, such as flooding rains or hot dry winds that spread wildfire.”

Using cBottle in NVIDIA Earth-2, developers can build climate digital twins to interactively explore and visualize kilometer-scale climate data, as well as predict possible scenarios at low latency and with high throughput.

The cBottle foundation model is available for early access. Climate AI researchers interested in retraining the model can access cBottle codebase from GitHub and the preprint on arXiv.

Watch the NVIDIA GTC Paris at VivaTech keynote from NVIDIA founder and CEO Jensen Huang, as well as the special address on NVIDIA CUDA-X libraries, to learn more.

See notice regarding software product information.