Tech giant IBM has announced a new open-source AI foundation model for different weather and climate use cases that is accessible to the business, developer, and scientific community. With contributions from Oak Ridge National Laboratory and developed by IBM and NASA, the model provides a flexible and scalable solution for various challenges related to both short-term weather forecasting and long-term climate projection.
According to a paper recently published on arXiv, “Prithvi WxC: Foundation Model for Weather and Climate,” the weather and climate foundation model can tackle far more applications than existing weather AI models because of its unique design and training regime.
The model’s potential applications include creating targeted forecasts based on local data, identifying and predicting severe weather patterns, improving the spatial resolution of global climate simulations, and enhancing how physical processes are represented in numerical weather and climate models.
In one experiment, outlined in the paper, the foundation model accurately reconstructed global surface temperatures using a random sample of just five percent original data, suggesting a wider application to issues with data assimilation.
NASA’s Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2) provided the 40 years of Earth observation data used to pre-train this model. Its distinctive architecture as a foundation model enables it to be adjusted to local, regional, and global scales. It is appropriate for a variety of weather studies because of its adaptability.
Along with two fine-tuned versions of the model addressing specific scientific and industry-relevant uses, the foundation model is accessible for download via Hugging Face. These are:
- Climate and weather data downscaling
- Gravity wave parameterization
Director of the Earth Science Division of NASA’s Science Mission Directorate, Karen St. Germain, said “Advancing NASA’s Earth science for the benefit of humanity means delivering actionable science in ways that are useful to people, organizations, and communities. The rapid changes we’re witnessing on our home planet demand this strategy to meet the urgency of the moment.”
“The NASA foundation model will help us produce a tool that people can use: weather, seasonal, and climate projections to help inform decisions on how to prepare, respond, and mitigate,” he added.
IBM has already collaborated with Environment and Climate Change Canada (ECCC) to evaluate the model’s versatility with new weather forecasting use cases.
Using precipitation nowcasting, a method that takes real-time radar data as input, ECCC is experimenting with very short-term precipitation forecasts with the model. The team is also testing the downscaling approach from global model forecasts at 15 km to km-scale resolution.
This weather and climate model joins the Prithvi family of AI foundation models and is a part of a larger collaboration between IBM Research and NASA to employ AI technology to investigate our world.
The Prithvi geospatial AI foundation model was made the most open-source geospatial AI model on Hugging Face last year by IBM and NASA. Since then, governments, businesses, and governmental organizations have examined changes in land use, biodiversity, disaster patterns, and other geophysical phenomena using this geospatial foundation model.
The IBM Granite Hugging Face page provides access to the downscaling model, while the NASA-IBM Hugging Face page offers access to the foundation and gravity wave parameterization models.
Stay Tuned to The Future Talk for more AI news and insights!