NVIDIA promotes open Earth-2 weather models across Europe with new StormScope EU pipeline

Weather can be a business risk long before it becomes news. NVIDIA says its “Earth-2” open model family is now being deployed across Europe to produce more locally relevant forecasts for tasks ranging from energy planning to insurance and transportation routing.

In a new post tied to GTC Berlin, NVIDIA describes Earth-2 as a set of open AI models and developer tools that can be customized with user data and run on NVIDIA computing. The company positions Earth-2 as a way for research teams and developers to build weather and climate applications that go beyond a single global view by tailoring outputs to specific regions.

The centerpiece for the European push is “StormScope EU,” which NVIDIA describes as a regional version of StormScope designed for short-term forecasting. NVIDIA says StormScope EU is trained on Europe weather and climate data and uses EUMETSAT satellite observations to predict evolving cloud conditions across 16 Meteosat channels in 10-minute intervals.

NVIDIA says it is showcasing a “live StormScope EU pipeline” in an interactive weather demo built with partner Earthmover. The demo, according to NVIDIA, uses Earth2Studio to run forecasts and Earthmover’s data platform to manage and serve results that are refreshed as new satellite observations arrive every 10 minutes.

Rather than requiring users to download full datasets, NVIDIA says Earthmover serves the outputs as map tiles, allowing demo visitors to explore predictions across Europe through a web-style map interface. The company’s pitch is that open models plus streaming-style delivery can make advanced weather AI easier to test and operationalize.

NVIDIA also highlights how developers can combine multiple Earth-2 components in custom workflows. The post says teams can use StormScope for short-term forecasting and use “CorrDiff” to add local detail to coarser weather data. It also notes that teams can take Earth-2 model architectures and train them for other environmental use cases, including air-quality modeling.

The company points to multiple organizations building on Earth-2 capabilities. AXA is using “FourCastNet 3” to study regional cyclone risk beyond historical records, generating insights across more than 1,000 simulated seasons to support risk assessment and prevention targeting. Vaisala Xweather, NVIDIA says, is combining Earth-2 with proprietary observations, lightning data and forecasting expertise to improve weather intelligence for government, energy, maritime, transportation and insurance customers, including adapting StormScope with lightning data to advance forecast lead time.

At the academic and energy-security end of the ecosystem, NVIDIA says researchers at the University of Manchester adapted the Earth-2 CorrDiff model for U.K. air-quality modeling. The team, according to NVIDIA, trained the model on the NVIDIA-powered Isambard-AI supercomputer and ran it locally on NVIDIA DGX Spark systems, aiming to bring environmental modeling from a national supercomputer to a workstation setup. TotalEnergies, NVIDIA adds, is exploring Earth-2 for solar-energy forecasting by evaluating StormScope-based solar-radiation nowcasts at two photovoltaic sites and developing models intended to provide weather information at subhourly intervals.

NVIDIA also describes other research applications in the same open-model ecosystem, including Space42’s “SARDiff4Height,” which NVIDIA says uses CorrDiff and PhysicsNeMo to estimate surface heights from a single radar image and generate uncertainty maps. NVIDIA says PhysicsNeMo is an open source framework that provides training and customization tools, and the company uses that to explain how teams can incorporate domain knowledge and proprietary data while building region-specific applications.

The company does not provide performance metrics in the post, such as forecast accuracy benchmarks, latency targets for operational deployment, or comparisons to existing commercial weather products. It also does not detail how partners integrate the Earth-2 models into production systems beyond the demo architecture, including how quality controls and verification would be handled in live environments.

Looking ahead, the key question for observers is whether the “open models plus near-real-time delivery” approach scales from demos to sustained operational use. With StormScope EU refreshed every 10 minutes in the showcased pipeline, the next proof point will likely be whether accuracy, reliability and ease of integration satisfy the validation standards required by weather-dependent industries like energy, insurance and logistics. NVIDIA’s Earth-2 strategy will be tested as more regional workflows move from research settings into ongoing production.