NVIDIA Earth-2

Earth-2 is a digital twin platform for producing accurate and dynamic weather data for weather prediction. Earth-2 makes use of NVIDIA’s CorrDiff, a Generative AI model that provides higher resolution data than most current models

Earth-2 leverages NVIDIA’s Omniverse platform, which allows customers to develop 3D workflows and applications. It enables OpenUSD applications and allows them to better visualize and intuitively understand weather conditions

The Need for Digital Twins in Weather Prediction

The impact of extreme climate events are costing hundreds of billions of dollars each year

- Dion Harris, Head of Datacenter Product Marketing, NVIDIA

  • We don’t have enough realistic data to train existing AI weather models to perform accurate and dynamic predictions of weather amid a changing climate. Match the intensity and pace of weather
  • Weather is become more disruptive. This requires more operational planning and prediction reliability. Current models make use of synthetic weather data, which may not be accurate and reliable all the time

TWC Digital Twin Technology

TWC has a proprietary physics-based weather-forecasting model called GRAF, in collaboration with NVIDIA and National Center for Atmospheric Research (NCAR). GRAF is rapidly updating and predicts more realistic and small-scale detail than most conventional weather models

TWC’s Weatherverse is a digital twin simulation that uses real datasets (historical, current, and forecasted) at a global scale. Weatherverse will be made available on NVIDIA’s Omniverse

Added Value and Benefits

  • Customers building digital twins can better assess and visualize the actual effects of weather in an intuitive manner because the Weatherverse digital twin is driven by real datasets

Relevant Links