Research team releases new artificial intelligence model for weather and climate applications

Working together, NASA and IBM Research have developed a new artificial intelligence model to support a variety of weather and climate applications. The new model – known as the Privthi-Weather-Climate Core Model – uses artificial intelligence (AI) in a way that could significantly improve the resolution we will be able to obtain, opening the door to better regional and local weather and climate models.

Baseline models are large-scale base models that are trained on large, unlabeled datasets and can be tuned for a variety of applications. The Privthi-weather-climate model is trained on a broad data set – in this case, NASA’s Modern Era Retrospective Analysis for Research and Applications (MERRA-2) data – and then uses AI learning capabilities to apply the collected patterns to based on initial data in a wide range of additional scenarios.

“NASA’s advancement of 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 are witnessing on our home planet require this strategy to meet the urgency of the moment.” said Karen St. Germain, director of the Earth Sciences Division in NASA’s Science Mission Directorate. “NASA’s core model will help us create a tool that people can use: weather, seasonality and climate forecasts that will help make decisions about how to prepare, respond and mitigate.”

With the Privthi-Weather-Climate model, researchers will be able to support a wide variety of climate applications that can be leveraged across the scientific community. These applications include detecting and predicting severe weather or natural disasters, creating targeted forecasts based on local observations, improving the spatial resolution of global climate simulations down to regional levels, and improving the representation of how physical processes are incorporated into weather and climate models.

“These revolutionary AI models are transforming data accessibility, significantly lowering the barrier to accessing NASA science data,” said Kevin Murphy, NASA chief science data officer in the Science Mission Directorate at NASA Headquarters. “Our open approach to sharing these models encourages the global community to discover and leverage the capabilities we have developed, ensuring that NASA’s investment enriches and benefits all.”

Privthi-weather-climate was developed through open collaboration with IBM Research, Oak Ridge National Laboratory, and NASA, including the Interagency Implementation and Advanced Concepts Team (IMPACT) at the Marshall Space Flight Center in Huntsville, Alabama.

Due to the flexibility of the model architecture, Privthi-weather-climate can capture the complex dynamics of atmospheric physics even when information is missing. This basic weather and climate model can be scaled to both global and regional areas without loss of resolution.

“This model is part of our overall strategy to develop a family of core AI models that will support NASA’s science mission objectives,” said Rahul Ramachandran, IMPACT project leader at Marshall. “These models will enhance our ability to draw conclusions from our extensive Earth observation archives.”

Privthi-weather-climate is part of a larger family of models – the Privthi family – which includes models trained on NASA Harmonized LandSat and Sentinel-2 data. The latest model is openly collaborative in accordance with NASA’s Open Science Principles to ensure all data is accessible and useful to communities around the world. It will be released later this year on Hugging Face, a machine learning and data science platform that helps users build, deploy and train machine learning models.

“The development of NASA’s core weather and climate model is an important step toward democratizing NASA’s science and observation mission,” said Tsendgar Lee, program manager for Weather Research and Analysis, High-End Computing Program and Data, NASA Operations and Evaluation. “We will continue to develop new technology for climate scenario analysis and decision-making.”

Along with IMPACT and IBM Research, significant contributions to the development of Privthi-Weather-Climate included NASA’s Office of the Chief Science Data Officer, NASA’s Global Modeling and Assimilation Office at Goddard Space Flight Center, Oak Ridge National Laboratory, University of Alabama at Huntsville, U.S. Colorado State and Stanford University.