IBM and NASA have joined forces to develop a groundbreaking Earth sciences model, referred to as the GPT of Earth sciences. NASA predicts that its Earth science missions will generate an enormous amount of data, reaching approximately a quarter million terabytes in 2024. To effectively analyze this vast volume of raw satellite data, IBM, in collaboration with HuggingFace and NASA, has worked on an open-source geospatial foundation model. This model will lay the groundwork for a new generation of climate and Earth science artificial intelligence systems capable of various tasks, including deforestation monitoring, crop yield prediction, and tracking greenhouse gas emissions.
IBM employed its recently-launched Watsonx.ai as the foundational model, utilizing a year’s worth of NASA’s Harmonized Landsat Sentinel-2 satellite data (HLS). The HLS data is collected by the ESA’s pair of Sentinel-2 satellites, which are specifically designed to capture high-resolution optical imagery over land and coastal regions using 13 spectral bands.
HuggingFace is hosting the model on its open-source AI platform. Through fine-tuning the model on labeled data for tasks like flood and burn scar mapping, the team managed to achieve a 15 percent performance improvement compared to the current state-of-the-art models while using only half as much data.
Sriram Raghavan, VP of IBM Research AI, highlighted the critical role of open-source technologies in advancing areas like climate change research. Combining IBM’s efforts to create adaptable and reusable AI systems with NASA’s repository of Earth-satellite data, and making it available on Hugging Face’s leading open-source AI platform, facilitates collaborative efforts to develop faster and more impactful solutions for the benefit of our planet.
Frequently Asked Questions (FAQs) about Earth sciences model
What is the GPT of Earth sciences?
The GPT of Earth sciences is an open-source geospatial foundation model developed through a collaboration between IBM, NASA, and HuggingFace. It serves as the basis for a new class of climate and Earth science AIs.
How much data is generated by NASA’s Earth science missions in 2024?
NASA estimates that its Earth science missions will generate around a quarter million terabytes of data in 2024 alone.
What tasks can the Earth sciences model perform?
The model can track deforestation, predict crop yields, and monitor greenhouse gas emissions, among other tasks related to Earth sciences.
How was the model created?
IBM leveraged its Watsonx.ai as the foundational model, using a year’s worth of NASA’s Harmonized Landsat Sentinel-2 satellite data (HLS).
What is the role of HuggingFace in this collaboration?
HuggingFace is hosting the model on its open-source AI platform, making it accessible to the research community and fostering collaboration.
How much did fine-tuning the model improve its performance?
By fine-tuning the model on labeled data for flood and burn scar mapping, the team achieved a 15 percent performance improvement over the current state-of-the-art models while using half as much data.
Why is open-source technology crucial for climate change research?
Open-source technologies like this Earth sciences model accelerate critical areas of discovery, enabling faster and more impactful solutions for improving our planet.
More about Earth sciences model
- IBM Research: https://www.research.ibm.com/artificial-intelligence/watsonx/
- NASA Earth Science: https://science.nasa.gov/earth-science
- HuggingFace: https://huggingface.co/
- ESA Sentinel-2: https://sentinel.esa.int/web/sentinel/missions/sentinel-2
5 comments
wow, so much data, quarter mil terabytes in 2024 alone! ibm’s Watsonx.ai used with NASA data, hugginface hostin model. amazin collaboration, open-source is the future!
fine-tuned the model, 15% improvement, half data! dat’s somethin! earth sciences AI gonna help our planet, yay! hugginface platform makes it accessible, kudos!
open-source model, earth sciences, bringin researchers together! faster solutions for climate change, need more of these collaborations! go IBM, NASA, hugginface!
deforestation, crop prediction, greenhouse gases, all in one model? amazin! open-source tech accelerates climate change research. IBM and NASA rockin it!
nasa and ibm workin together, buildin a gpt for earth science, thatz cool! so much data to help climate science, predictin crops and greenhouse gases, awesome!!