IBM, NASA Will Use AI to Improve Climate Change Research
NurPhoto /Getty
The research will use IBM’s artificial intelligence technology and NASA’s Earth and geospatial science data for easier and faster research.
IBM is collaborating with NASA’s Marshall Space Flight Center to use artificial intelligence for climate change research in an effort to make research analysis of these large datasets easier and faster, according to an announcement Wednesday.
The research will use IBM’s AI foundation model technology and NASA’s Earth and geospatial science data, specifically NASA’s Earth-observing satellite data. The collaboration will help “provide an easier way for researchers to analyze and draw insights from these large datasets.”
As noted in the announcement, foundation models—types of AI models—“are trained on a broad set of unlabeled data,” utilized for different tasks and can “apply information about one situation to another.” Foundation models have furthered natural language processing technology and IBM is working on foundation model applications besides language, according to the announcement.
Meanwhile, the increasing amount of data from Earth observation studies and monitoring can provide a valuable research resource. According to the announcement, IBM’s foundation model technology could help make the discovery and analysis of this data faster to rapidly advance scientific understanding of Earth and climate change issues.
“Foundation models have proven successful in natural language processing and it’s time to expand that to new domains and modalities important for business and society,” Raghu Ganti, principal researcher at IBM, said. “Applying foundation models to geospatial, event-sequence, time-series and other non-language factors within Earth science data could make enormously valuable insights and information suddenly available to a much wider group of researchers, businesses and citizens. Ultimately, it could facilitate a larger number of people working on some of our most pressing climate issues.”
As part of this collaboration, IBM and NASA will create several new technologies to gain insights from Earth observations. Specifically, they will train an IBM geospatial intelligence foundation model on NASA’s Harmonized Landsat Sentinel-2 dataset, which provides a record of land cover and use changes as seen by satellites. Researchers will be able to study Earth’s systems by analyzing this data with the AI technology to identify changes in natural disasters, cyclical crop yields and wildlife habitats, among other things.
The collaboration will also help create an easy-to-search collection of Earth science literature as IBM has created a natural language processing model trained from approximately 300,000 Earth science journal articles “to organize the literature and make it easier to discover new knowledge.” The model uses PrimeQA—IBM’s open-source multilingual question-answering system. This language model could be incorporated into NASA’s scientific data management and stewardship processes, the announcement stated.
“The beauty of foundation models is they can potentially be used for many downstream applications,” Rahul Ramachandran, senior research scientist at NASA’s Marshall Space Flight Center, said. “Building these foundation models cannot be tackled by small teams. You need teams across different organizations to bring their different perspectives, resources and skill sets.”
IBM and NASA would also establish a foundation model for weather and climate prediction by using atmospheric observation dataset, MERRA-2. This effort is part of NASA’s Open-Source Science Initiative, which seeks to build an inclusive, transparent and collaborative open science community over the next ten years.
The collaboration between IBM and NASA comes after the agency has worked on other technologies to improve Earth and climate-related observations. For example, the new Surface Water and Ocean Topography satellite—as part of a project with France’s space agency—will observe Earth’s water. Moreover, NASA space satellites were also used to pinpoint CO2 emission at a source level.