Wed Sep 25 07:25:05 UTC 2024: ## NASA and IBM Release Open-Source AI Climate Model for More Efficient Weather Forecasting

**[City, State] -** Researchers at IBM and NASA have teamed up to develop an open-source AI climate model named “Prithvi WxC” designed to predict weather patterns accurately while utilizing significantly fewer computing resources than traditional physics-based models. This foundation model, trained on 40 years of NASA’s MERRA-2 dataset, boasts 2.3 billion parameters and has proven capable of generating accurate global surface temperatures using only 5% of the original data.

Prithvi WxC’s unique strength lies in its adaptability. Unlike existing AI models focused on specific datasets and use cases, this foundation model can be tailored to a variety of applications, ranging from short-term weather forecasting to long-term climate projections.

“We have designed our weather and climate foundation model to go beyond such limitations so that it can be tuned to a variety of inputs and uses,” stated Juan Bernabe-Moreno, director of IBM Research Europe.

The model’s release on Hugging Face includes two fine-tuned models designed for climate and weather downscaling and gravity wave parameterization, providing researchers with valuable tools for improving weather warnings, climate simulations, and understanding atmospheric processes.

“The NASA foundation model will help us produce a tool that people can use [for] weather, season and climate projections to help inform decisions on how to prepare, respond, and mitigate,” said Karen St Germain, director of NASA’s Earth Science Division.

The model’s relatively small size, trained on a cluster of 64 Nvidia A100s, makes it accessible for various climate centers with GPU-equipped supercomputing clusters.

The Canadian government has already adopted Prithvi WxC for weather forecasting, particularly focusing on very short-term precipitation forecasts and downscaling for more localized predictions. This collaboration highlights the model’s potential to revolutionize climate modeling and improve weather forecasting capabilities worldwide.

Read More