By ToAI Team
The need to advance climate research has become more evident as global warming escalates and extreme weather events all around the world increase. Since such a need has been realized, NASA has not only increased its investigations but has also grown the capacity of external researchers. A major step in this direction is the use of artificial intelligence (AI), which NASA considers an essential tool for achieving climate study goals.
In a joint pursuit for 2023, NASA collaborated with IBM Research to build an AI geospatial foundation model. NASA’s colossal Harmonized Landsat and Sentinel-2 (HLS) dataset, a contributing resource, trains the innovative model under construction, providing a general AI research base. Indicative of the “open science” paradigm, NASA has shared the model, which is free for any interested party.
The foundation models are the very definition of the role of a Swiss Army Knife for data science, claimed Manil Maskey, data science lead at NASA’s Office of the Chief Science Data Officer. Such models serve as guiding frameworks for scientists to develop customized approaches that help them cope with various environmental issues. To explain, the NASA team, which has been working with the Interagency Implementation and Advanced Concept Team (IMPACT), has already worked towards the precise creation of the model for several cases, including defining burn scars, mapping flood waters, and classifying varieties of land use.
The creation of such a powerful foundation model had a great demand for computing resources that was only possible with the support of a partnership. NASA offered its rich data and competence in science, while IBM made sure the computing infrastructure was adequate and improved AI techniques. With this collaboration, the model not only moved forward, but also exhibited the concept of open science, ensuring that the gains are not only limited to the research community but are inclusive of all.
Maskey emphasized the collaborative nature of the initiative and demanded the open availability of everything, including the training data, algorithms, and models. This transparency enables researchers to fully understand and use it.
NASA and IBM Research are now extending their partnership and aiming to construct a new base model to enhance weather and climate studies, driven by the significant success of the geospatial model. The project involves collaboration with Oak Ridge National Laboratory, NVIDIA, and a number of academic institutions. The next model will rely on the Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2) dataset, which contains comprehensive atmospheric data from 1980 to the present.
The current development is aimed at particular models that specialize in the various regions of the planet, with the perspective that, in the future, these could be combined into one comprehensive model. This exact “digital twin” of Earth could reinvent all climate and environmental analysis and forecasting.
The current efforts of NASA and IBM, such as the development of AI models and dedication to open science, reflect the huge potential of AI that, when applied to open science, will make a difference in the study of the earth. By democratizing access to these state-of-the-art tools, they are laying a foundation that can profoundly impact our knowledge of and stewardship of the world.
What is your reaction?
