April 16, 2024

NASA works with IBM on foundational data model for Earth science

4 min read



NASA’s Marshall Area Flight Heart has started perform with IBM to use synthetic intelligence (AI) technological know-how to learn new insights in its Earth and geospatial science knowledge. The joint hard work will utilize AI foundation product technologies to NASA’s Earth-observing satellite info for the initially time.

These observations are utilized by researchers to examine and keep an eye on Earth. NASA mentioned the knowledge is remaining gathered unparalleled prices and volume, consequently it wants new techniques to extract awareness from these rising details means. The pair prepare to acquire an less complicated way for researchers to analyse and attract insights from the significant datasets, dependent on IBM’s foundation model technological know-how.

According to IBM, foundation products are kinds of AI styles that are properly trained on a wide set of unlabelled information, can be utilised for distinct duties, and can implement information and facts about 1 problem to yet another. These products have fast state-of-the-art the industry of normal language processing (NLP) engineering about the past five decades, and IBM is pioneering apps of foundation designs further than language.

“The attractiveness of basis models is they can potentially be employed for several downstream apps,” claimed Rahul Ramachandran, senior research scientist at NASA’s Marshall Room Flight Heart in Huntsville, Alabama. “Building these foundation designs cannot be tackled by compact teams. You will need groups across different organisations to deliver their distinct perspectives, sources and skillsets.” 

IBM and NASA strategy to develop numerous new technologies to extract insights from Earth observations. A single venture will teach an IBM geospatial intelligence foundation product on NASA’s Harmonized Landset-Sentinel-2 (HLS) dataset, a history of land cover and land use variations captured by Earth-orbiting satellites.

By analysing petabytes of satellite info to recognize adjustments in the geographic footprint of phenomena these as normal disasters, cyclical crop yields and wildlife habitats, IBM and NASA hope this foundation model technologies will aid researchers present essential examination of our planet’s environmental units.

NASA will also be doing the job with IBM to make Earth science literature very easily searchable. IBM has produced a all-natural language processing (NLP) product properly trained on just about 300,000 Earth science journal articles or blog posts to organise the literature and make it less difficult to find out new expertise. According to IBM, this signifies the most significant AI workload skilled on Purple Hat’s OpenShift application. The totally experienced product utilizes PrimeQA, IBM’s open resource multilingual issue-answering process. Beyond providing a resource to scientists, the new language product for Earth science will be built into NASA’s scientific knowledge management and stewardship processes.

One more opportunity IBM-NASA joint project is the construction of a basis design for temperature and weather prediction employing MERRA2, a dataset of atmospheric observations. This collaboration is component of NASA’s Open Source Science Initiative, a dedication to constructing an inclusive, transparent and collaborative open science group over the next ten years.


Supply hyperlink NASA and IBM have joined forces to develop a data model to store and share Earth science data.

NASA’s Earth Science Data and Information System (ESDIS) is responsible for taking measurements from satellites and other scientific instruments, and for sharing these data with the public. The ESDIS team is working with IBM to develop a foundational data model that will make it easier for researchers to collaboratively use Earth science data from around the world. This data model will be based on open standards, such as the World Wide Web Consortium’s Resource Description Framework (RDF).

The RDF data model will provide a standard way of representing and sharing Earth science data, across different traditional data formats, hardware platforms, and computing systems. This will make it easier for researchers to collaborate, analyse, and share data across text, voice, images, and more.

In addition to providing an easy-to-use data model for Earth science data, the collaboration between NASA and IBM also aims to provide a unified platform for the development of analytics and applications that can be applicable to many different areas of Earth science. This includes the development of machine learning models, analytics tools, and applications for climate change and land management.

NASA Deputy Administrator Dava Newman said: “Earth science data can provide valuable insights into weather, land use, climate change, and a range of other topics. Our collaboration with IBM allows us to leverage the power of data to build a new understanding of our planet and create opportunities for economic growth.”

The data model developed by NASA and IBM will benefit researchers and the public alike. It will be openly available to researchers and organisations interested in using Earth science data, and will help foster innovation and collaboration in the field. By making Earth science data easier to access and use, scientists and organisations can answer vital Earth science questions more efficiently.

The collaboration between NASA and IBM marks a new era in the field of Earth science. By leveraging the power of technology and open standards, they have made it possible to easily access and use data from the planet to better understand our environment.