Facing the challenge of increasing extreme weather events worldwide, the National Aeronautics and Space Administration (NASA) has partnered with IBM Research to develop a new AI geographic spatial foundational model based on artificial intelligence and open science principles.
According to recent announcements from NASA, the agency collaborated with IBM Research to develop a new AI geographic spatial foundational model based on artificial intelligence and open science principles. This model is trained using NASA Harmonized Landsat and Sentinel-2 data (HLS), providing not only more data sources and support for scientific research but also enhancing the ability to forecast extreme weather events in advance.
Manil Maskey, the data science lead at NASA’s Office of the Chief Science Data Officer (OCSDO), described this model as “fundamental in understanding what is present in the data; it’s like a Swiss Army knife, capable of being used in various situations and handling different tasks.” This versatility means that once the foundational model is created, it can be trained on small amounts of data to perform specific tasks, thus achieving robust and efficient solutions.
So far, the Interagency Implementation and Advanced Concept Team (IMPACT) and its partners have demonstrated the potential of this geographic spatial foundational model in multiple fields, including monitoring forest fires, delineating flood zones, and classifying crops and other land use types.
Given that creating the initial foundational model requires vast amounts of data and powerful computing capabilities, the collaboration between NASA and IBM is crucial. In this project, NASA provided data and scientific expertise, while IBM contributed computational power and expertise in optimizing AI algorithms. This collaboration not only accelerates the development process of the model but also ensures that research outcomes benefit more researchers through the principles of open science.
Manil Maskey pointed out, “To build a large-scale foundational model, we realized early on that success wouldn’t come from the efforts of a single institution alone. This is why we keep the foundational model open to the public, maintaining transparency from pre-training data, code, to practices, model weights, fine-tuning training data, and publications, allowing researchers to trace the reasons for using certain data or model architectures.”
Building on the success of the geographic spatial foundational model by NASA and IBM Research, the two entities continue to collaborate on creating a new model for weather and climate research. Other collaborators in this new project include Oak Ridge National Laboratory (ORNL), NVIDIA, and several universities. The primary dataset for this new weather and climate model is the Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2), a vast atmospheric reanalysis dataset from 1980 to the present. Similar to the geographic spatial foundational model, this new weather and climate model will also be developed using open science methods and will be made available to the public in the near future.
