Nvidia expands Earth-2 family with three new open source models for advanced weather and climate simulations

by
Alyssa Davidson

Posted: January 27, 2026 5:03 AM Updated: January 27, 2026 5:03 AM

by well

Edit and fact check date: January 27, 2026, 5:03 a.m.

We sometimes use automatic translation plugins to improve your local language experience. The automatic translation may not be accurate, so please read the original text for accurate information.

Technology company NVIDIA has announced three new open source models within its Earth-2 family that aim to streamline weather forecast application development across the entire meteorological workflow, including data assimilation, forecasting, current casting, downscaling, and related tasks.

Developers can quickly construct weather and climate simulations by leveraging NVIDIA’s open source software, such as Earth2Studio for building inference pipelines and Physics Nemo for model training.

The Earth-2 suite integrates a variety of accelerated tools and models to unify capabilities that have traditionally been scattered across weather and climate AI environments. The fully open framework allows developers to customize and enhance simulations to their own needs by leveraging proprietary data and infrastructure to create self-managed weather and climate forecasts.

This approach allows Earth-2 to provide a leading set of open weather and climate models, supported by an accessible software ecosystem, to generate fully governed, custom forecast capabilities.

NVIDIA Expands Earth-2 Weather AI Suite to Hug Face with StormScope, Atlas, and HealDA Models

Earth-2 Nowcasting, released by Hugging Face, introduces a model architecture called StormScope that leverages generative AI to produce high-resolution near-term forecasts. The system generates kilometer-scale forecasts for local storms and hazardous weather up to six hours in advance, delivering results within minutes.

By directly simulating storm dynamics and predicting satellite and radar output, Earth-2 Nowcasting can outperform existing physics-based models in short-term precipitation forecasts. The current version is trained on geostationary satellite observations over the contiguous United States, but the methodology can be applied to other regions with similar satellite coverage.

Earth-2 Medium Range, also available on the platform at the same time, is built on the new Atlas model architecture designed for medium-range forecasts up to 15 days. The system predicts more than 70 atmospheric variables, including temperature, wind, pressure, and humidity, using a latent diffusion transducer architecture that models gradual changes in the atmosphere to maintain structural consistency and minimize errors. Benchmark tests show that Earth-2 Medium Range outperforms other public forecast models, including GenCast, on commonly used industry variables.

Additionally, Hugging Face’s upcoming Earth-2 Global Data Assimilation features the HealDA architecture, which generates initial atmospheric conditions for thousands of locations around the world, including temperature, wind speed, humidity, and barometric pressure. By generating these snapshots in seconds on GPUs rather than hours on traditional supercomputers, HealDA can initialize predictions faster and more accurately. Combined with the Earth-2 Medium Range, this creates a complete AI-based pipeline capable of delivering highly accurate forecasts at scale.

These additions expand NVIDIA’s existing suite of open weather and climate models, including FourcastNet3, CorrDiff, cBottle, and DLESym. NVIDIA Earth2Studio, which supports this ecosystem, provides an open-source Python framework for rapidly building AI weather and climate simulations and provides tools for inference and integration with new model checkpoints.

As MPost’s resident journalist, Alisa specializes in the broad areas of cryptocurrencies, zero-knowledge proofs, investing, and Web3. With a keen eye for new trends and technologies, she provides comprehensive coverage to inform and engage readers about the ever-evolving digital financial landscape.

• 
  
  

more articles