NASA and IBM (NYQ: IBM) have teamed up to combine the former’s geospatial data with the latter’s foundation artificial intelligence (AI) models to study climate change. The collaboration includes a future project that is part of NASA’s Open-Source Science Initiative to help scientific research at the nascent stage, which could benefit climate-tech startups.
- IBM will apply its AI technology to vast amounts of data generated by NASA’s satellites to study the Earth and climate-related issues.
- The partnership will provide researchers with a deeper understanding of environmental issues such as land use changes and natural disasters.
- A future project planned by IBM and NASA may also benefit early-stage climate-tech startups as it will be part of an open-source initiative.
AI is the perfect companion technology to the vast amounts of data generated by the information and images captured by satellites, which can be used to develop actionable insights. The number of applications addressing climate change that combine the two technologies is on the rise. Examples include capturing thermal-infrared images to detect and even predict wildfires, and using satellite monitoring to verify the effectiveness of carbon removal projects.
Satellites can also gather data to monitor infrastructure, including energy infrastructure, which goes beyond the data capabilities of traditional ground and aerial surveillance methods. In addition to providing access to hard-to-reach places, using satellites lowers the emissions from the operation and maintenance of manual methods.
IBM and NASA collaborate to tackle climate change
IBM is teaming up with NASA’s Marshall Space Flight Centre to combine its AI technology with NASA’s large repository of earth and geospatial science data. Satellite data enables the study of the planet from different aspects, and for different applications. AI will sift through the vast amount of data being generated, extracting the relevant information to provide actionable insights.
For each of these applications, the data needs to be tagged to suit the specific task at hand, in order to be recognised by the AI model. This can take up thousands of man-hours and computing time depending on the type of data used.
According to IBM, the computing power and time needed to train an AI model capable of recognising natural language, for example, has about the same carbon footprint as the lifetime emissions of five cars. To address this challenge, the scientists at IBM have developed reusable, open-source-based models that can form the foundation of any application with small modifications.
What are foundation models and how can they help with climate-related issues?
IBM’s foundation model technology developed is intended to feasibly deploy multiple AI solutions aimed at climate change. The company claims it provides the speed and granularity needed to process and analyse large amounts of data.
A key distinguishing feature of foundation models is that they can work with unlabelled data. Early examples of foundation model AI have shown success in the areas of imagery and language, such as ChatGPT, which can create original text or content based on a user prompt or input.
By training itself, using machine learning techniques such as self-supervised learning and transfer learning, the model can use the same information across multiple applications, which makes it the foundation for multiple use cases. In its collaboration with NASA, IBM is applying this technology to help scientists understand the multiple issues that are causing climate change, providing the analysis that can help devise solutions.
What projects have IBM and NASA planned?
IBM and NASA plan to gather insights into the physical, chemical and biological features of the planet using satellites and remote-sensing technologies, or Earth observations.
The first project involves training IBM’s geospatial intelligence foundation model on NASA’s Harmonized Landsat Sentinel-2 dataset, which will provide critical information about the impact of climate disasters, wildlife behaviours, and farming and land use, from petabytes, or millions of gigabytes, of satellite data.
The pair also plans to use a natural language processing model developed by IBM to enable search and retrieval from existing literature on Earth science. In addition to being a research resource, the model could be used by NASA’s stewardship processes, such as those used to develop its Mission Algorithm and Analysis Platform.
“The beauty of foundation models is they can potentially be used for many downstream applications,” said Rahul Ramachandran, senior research scientist at NASA’s Marshall Space Flight Center in Huntsville, Alabama.
Raghu Ganti, principal researcher at IBM, added: “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.”
What is the future of the IBM-NASA collaboration?
A second project that has been planned by the collaboration takes advantage of NASA’s Open-Source Science Initiative. It will use an atmospheric observation dataset known as MERRA-2, to build a foundation model for weather and climate prediction.
The Open-Source Science initiative is NASA’s plan to build an open and inclusive science community by sharing software, data and research to help scientific research in its nascent stage. It has also designated 2023 as ‘The Year of Open Science’, during which it plans a series of events and activities to promote open engagement on scientific topics.
Collaborations such as the one between IBM and NASA, alongside initiatives such as the Open-Sources Science Initiative, can provide open access to technology which could benefit many early-stage climate-tech companies that are addressing the impacts of climate change.
