Pioneers Across Europe Deploy NVIDIA Earth-2 Open Models to Power Real-World Weather and Climate Prediction
NVIDIA Earth-2 open AI models enable customized weather and climate forecasting, with StormScope EU providing 10-minute interval cloud predictions across Europe using EUMETSAT data.
Weather forecasts influence critical infrastructure decisions, requiring both global storm tracking and localized insights for energy, transport and insurance sectors. NVIDIA Earth-2 offers open AI models and tools that developers can adapt with proprietary data to generate tailored forecasts. The platform supports global, regional and environmental modeling, bridging large-scale predictions with local operational needs. StormScope EU joins the Earth-2 family as a regional model trained on European weather data, delivering high-frequency cloud forecasts using satellite observations.
At GTC Berlin, NVIDIA demonstrated a live StormScope EU pipeline in an interactive demo built with Earthmover, showcasing real-time updates every 10 minutes. The system uses Earth2Studio for forecasting and Earthmover’s platform to serve results as map tiles, allowing users to explore predictions without downloading full datasets. Developers can integrate StormScope for short-term forecasting with CorrDiff to enhance local detail in coarser weather data. The open-source PhysicsNeMo framework enables teams to customize models for proprietary data and new applications like air-quality modeling.
AXA employs NVIDIA Earth-2’s FourCastNet 3 to simulate over 1,000 cyclone seasons, extending risk assessments beyond historical records to inform prevention strategies. Vaisala Xweather combines Earth-2 with its lightning data and proprietary observations to improve short-term forecasting lead times for government, energy and insurance customers. The company integrates StormScope with lightning data and offers these insights through its software-as-a-service platforms and AI assistants.
Researchers at the University of Manchester adapted CorrDiff for U.K. air-quality modeling using the Isambard-AI supercomputer and NVIDIA DGX Spark systems. TotalEnergies is testing StormScope-based solar-radiation nowcasts at two photovoltaic sites to refine solar-energy forecasting at subhourly intervals. Space42 developed SARDiff4Height using CorrDiff and PhysicsNeMo to estimate surface heights from radar images and generate uncertainty maps for closer analysis.