Nvidia's next-gen full Earth-2 digital twin AI models will predict and forecast the next big storm faster and more accurately (www.techradar.com)

Nvidia has launched its Earth-2 suite of AI models designed to revolutionize weather forecasting by significantly enhancing speed and accuracy while drastically reducing computational costs. Key components of Earth-2 include innovations like CorrDiff, which downscales rough continental predictions into high-resolution forecasts up to 500 times faster than traditional methods, and FourCastNet3, which delivers superior forecasts for wind, temperature, and humidity with a speed increase of up to 60 times. The integration of AI with high-resolution satellite and radar data enables continuous and actionable insights, crucial for various sectors including energy and disaster response. The importance of Earth-2 extends beyond mere speed; its open-source availability on platforms like Hugging Face and GitHub democratizes access to these advanced forecasting tools, allowing researchers and organizations to tailor models to specific local conditions. By leveraging simplistically designed scalable architectures, Nvidia aims to provide fundamental building blocks applicable to national meteorological services, energy companies, and more, enhancing decision-making in critical areas like agriculture and risk assessment. As the Israel Meteorological Service and companies like TotalEnergies begin deploying these models, the potential for transforming operational standards in weather prediction becomes evident.