NVIDIA Earth-2 Open AI Models Bring Local Weather Forecasting to Europe and Beyond
Weather forecasts shape daily plans, but timely, detailed weather information is also critical for energy producers, insurance companies and transportation operators. Global forecasts can show the broader shape of a storm, while decisions about solar farms, transportation routes and insured properties depend on what happens locally.
NVIDIA Earth-2 provides developers and researchers with a family of open AI models and tools that can be customized with their own data. Running on NVIDIA compute, these technologies support global weather forecasting, detailed regional prediction and environmental research.
StormScope EU Enables High-Resolution European Weather Nowcasting
StormScope EU is the latest model to join Earth-2. This regional version of the StormScope model is designed for short-term forecasting and trained on European weather and climate data. Using EUMETSAT satellite observations, it predicts changes in cloud conditions across 16 Meteosat channels at 10-minute intervals.
At GTC Berlin, NVIDIA is showcasing a live StormScope EU pipeline in an interactive weather demonstration built in partnership with Earthmover. The demonstration uses Earth2Studio to generate predictions and Earthmover’s data platform to manage and deliver real-time results.
The pipeline updates every 10 minutes as new satellite observations arrive. Earthmover delivers the results as map tiles, allowing visitors to explore forecasts across Europe without downloading the entire dataset.
Customizable AI Models for Local Weather and Environmental Applications
Developers can combine Earth-2 models in workflows tailored to local needs. StormScope supports short-term forecasting, while CorrDiff can add local detail to coarse weather data. Developers can also adapt these model architectures for new environmental applications, including air quality modeling.
The open-source NVIDIA PhysicsNeMo framework provides training and customization tools that help teams build applications using their own data and domain expertise.
Organizations Applying NVIDIA Earth-2 to Real-World Challenges
AXA: Assessing Regional Cyclone Risk
AXA uses FourCastNet 3 on NVIDIA Earth-2 to study regional cyclone risk beyond historical records across more than 1,000 simulated seasons. The resulting insights inform risk assessments and help customers target prevention efforts.
Vaisala Xweather: Delivering Earlier Severe Weather Warnings
Vaisala Xweather combines Earth-2 with its unique observations, lightning data and forecasting expertise to power weather intelligence platforms for government, energy, maritime, transportation and insurance customers.
The effort includes adapting StormScope to lightning data to reduce lead times for short-term forecasts. These warnings are made available through Vaisala Xweather’s software-as-a-service offering and AI assistant integration.
Video provided by Vaisala Xweather.
University of Manchester: Modeling UK Air Quality
Researchers at the University of Manchester adapted the Earth-2 CorrDiff model for UK air quality modeling. They trained the model on an NVIDIA-powered Isambard-AI supercomputer and ran it locally on an NVIDIA DGX Spark system, bringing environmental modeling from the nation’s supercomputer to the desktop.
TotalEnergies: Improving Solar Energy Forecasting
TotalEnergies is researching NVIDIA Earth-2 for solar energy forecasting. The company is evaluating StormScope-based solar radiation nowcasts at two solar power plants and developing a model that provides weather information at sub-hourly intervals.
The study investigates how more detailed forecasting can help predict changes in solar power generation.
Space42: Estimating Surface Height from Radar Images
Space42 developed SARDiff4Height using CorrDiff and NVIDIA PhysicsNeMo. The study estimates surface height from a single radar image and creates an uncertainty map to help analysts identify areas where estimates need more scrutiny.
Explore NVIDIA Earth-2 Open Models and Tools
Explore NVIDIA Earth-2 open models and tools and the Earth2Studio forecast example. Read the AXA hurricane risk case study and learn how teams are adapting weather AI to address challenges in climate, air quality, renewable energy and environmental modeling.
Source: blogs.nvidia.com