- NASA partners with IBM Research to develop AI geospatial foundation model for climate research.
- Model trained on NASA’s extensive HLS data, freely accessible to all researchers.
- Foundation models enable diverse applications, refined through specialized training.
- The collaboration leverages NASA’s data and IBM’s computing expertise, anchored in open science principles.
- The next phase involves creating a weather and climate model using the MERRA-2 dataset.
- Continued commitment to open science ensures accessibility and transparency in research efforts.
Main AI News:
In an era marked by escalating climate-related crises worldwide, the imperative for intensified examination of our warming globe grows more pressing by the day. For NASA, this entails not only conducting in-depth analyses of these phenomena but also facilitating external researchers in doing the same. Spearheaded by the agency, artificial intelligence (AI) initiatives stand as a formidable resource in achieving these objectives.
In 2023, NASA embarked on a groundbreaking partnership with IBM Research, culminating in the development of an AI geospatial foundation model. This revolutionary model, meticulously trained on extensive datasets comprising NASA’s widely utilized Harmonized Landsat and Sentinel-2 (HLS) data, serves as a cornerstone for a myriad of AI-driven investigations targeting environmental challenges. Upholding the ethos of open science, this model is freely accessible to all.
Foundation models, functioning as pivotal frameworks, empower scientists to craft a diverse array of applications, thereby facilitating potent and streamlined solutions. Manil Maskey, the data science lead at NASA’s Office of the Chief Science Data Officer (OCSDO), elucidates, “Foundation models only know what things are represented in the data. It’s like a Swiss Army Knife—it can be used for multiple different things.”
Once established, a foundation model can undergo specialized training with minimal data input to execute a designated task. The Interagency Implementation and Advanced Concept Team (IMPACT), in collaboration with partners, has showcased the versatility of the geospatial foundation model by refining its capabilities to identify burn scars, delineate floodwaters, and classify land use categories, including crops.
Given the substantial computational resources requisite for the model’s inception, successful realization necessitated a synergistic alliance. Here, NASA contributed invaluable data and scientific acumen, while IBM provided indispensable computing prowess and AI algorithm optimization proficiency. Anchored in a shared commitment to transparency and accessibility through open science principles, the collaborative endeavor ensures the model’s utility to a broad spectrum of researchers.
Reflecting on the triumph of the geospatial foundation model, NASA and IBM Research are poised to extend their collaboration, embarking on the creation of a parallel model tailored for weather and climate studies. This ambitious undertaking involves partnerships with Oak Ridge National Laboratory (ORNL), NVIDIA, and multiple academic institutions. The primary dataset for this venture comprises the Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2), an extensive repository of atmospheric reanalysis data spanning four decades.
Aligned with the ethos of its predecessor, the weather and climate model is being nurtured within an open science framework and is slated for public release imminently. While comprehensive coverage of Earth science necessitates numerous foundation models tailored to distinct datasets, Maskey envisions a future where these models converge into a unified, comprehensive entity—a “digital twin” of Earth. Such an innovation promises unparalleled insights and prognostic capabilities for diverse climate and environmental phenomena.
Amidst the landscape of relentless innovation, NASA and IBM’s pioneering geospatial and climate foundation models stand as beacons of progress in Earth science. While AI augments the endeavors of researchers, their unwavering commitment to open science amplifies the horizons of exploration, democratizing access to cutting-edge tools and propelling groundbreaking research initiatives aimed at safeguarding our planet’s future.
Conclusion:
The collaboration between NASA and IBM in advancing AI-driven climate research marks a significant stride towards leveraging technology for environmental insights. By democratizing access to foundational models and embracing open science principles, this partnership not only fosters innovation but also sets a precedent for collaborative efforts in tackling pressing global challenges. For businesses, it underscores the importance of strategic partnerships and the integration of cutting-edge technologies in addressing sustainability issues.
