Key Takeaway
Nvidia’s Earth-2 is a cloud and GPU platform designed for AI-accelerated weather and climate digital twins. It offers development tools, microservices, and reference implementations for simulation and visualization. Key models include FourCastNet for atmospheric dynamics, CorrDiff for kilometre-scale guidance, and StormCast, a generative AI model for high-fidelity forecasts. CorrDiff is noted for being up to 1,000 times faster and 3,000 times more energy-efficient than traditional methods. Tom Hamill from The Weather Company highlights the significance of StormCast in addressing challenges in numerical weather prediction, expressing enthusiasm for collaboration with Nvidia on these deep learning models.
Nvidia’s Digital Twin Platform
Nvidia’s Earth-2 is a cloud and GPU platform designed for creating and operating AI-accelerated digital twins for weather and climate.
It offers development tools, microservices, and reference implementations for simulation and visualization.
The platform includes models such as FourCastNet for simulating atmospheric dynamics, CorrDiff for generating kilometer-scale guidance, and StormCast, a generative AI model for high-fidelity atmospheric dynamics.
Nvidia claims that its CorrDiff model operates up to 1,000 times faster and is 3,000 times more energy efficient than traditional methods for similar tasks.
“Producing computationally feasible storm-scale ensemble weather forecasts is one of the major challenges in numerical weather prediction,” said Tom Hamill, Head of Innovation at The Weather Company, in a statement to Nvidia.
“StormCast is a significant model that tackles these challenges, and The Weather Company is eager to collaborate with Nvidia on developing, evaluating, and potentially utilizing these deep learning forecast models.”
