NASA, IBM Create AI Model to Predict Weather

IBM, NASA and the Oak Ridge National Laboratory have launched a new open-source AI model for weather and climate applications they said goes beyond normal forecasting to predict weather patterns at a variety of scales.

The new model called Prithvi WxC has been pre-trained on 40 years of Earth observation data from NASA's Modern-Era Retrospective Analysis for Research and Applications, Version 2.

It was created using a novel design and training regime that enables it to tackle tasks beyond forecasting unlike many existing weather AI models, the organizations said. For example, it can be fine-tuned to global, regional and local scales, making it suited to a range of weather studies. 

“This space has seen the emergence of large AI models that focus on a fixed dataset and single use case, primarily forecasting,” said Juan Bernabe-Moreno, director of IBM Research Europe and IBM's accelerated discovery lead for climate and sustainability.

This new foundation model can be tuned to a variety of inputs and uses making it more versatile. 

“It can run both on the entire Earth as well as in a local context, meaning it is well-suited to help us understand meteorological phenomena such as hurricanes or atmospheric rivers, reason about future potential climate risks by increasing the resolution of climate models and finally inform our understanding of imminent severe weather events,” said Bernabe-Moreno.

Related:Nvidia Debuts AI-Powered Weather Forecasting Digital Twin

Potential applications include creating targeted forecasts based on local observations, detecting and predicting severe weather patterns, improving the spatial resolution of global climate simulations and improving how physical processes are represented in numerical weather and climate models.

In one experiment the foundation model accurately reconstructed global surface temperatures from a random sample of five percent original data, suggesting a broader application to problems in data assimilation. 

This example and a more detailed explanation of the model is outlined in a paper recently published on arXiv, called “Prithvi WxC: Foundation Model for Weather and Climate.”

Other notable applications of the model include climate and weather data downscaling, which is inferring high-resolution outputs from low-resolution variables. The model can depict both weather and climate data at up to 12x resolution, generating localized forecasts and climate projections. And more accurate gravity wave parameterization to understand how exactly these affect climate processes.

IBM has already collaborated with Environment and Climate Change Canada with a view to test the flexibility of the model with additional weather forecasting use cases. It’s also part of a larger collaboration between IBM Research and NASA to use AI technology to explore the planet.

Related:Google Hybrid AI Model Accurately Forecasts Weather

The model is available for download for scientific, developer and business communities on Hugging Face, along with two fine-tuned versions of the model that tackle specific scientific and industry-relevant applications.