NOAA’s National Environmental Satellite, Data, and Information Service (NESDIS) has published a study of emerging Artificial Intelligence-driven Earth Observation Digital Twin (EODT) technology authored by Lockheed Martin and NVIDIA under a Joint Ventures Partnership (JVP) Broad Agency Announcement contract awarded in September 2022.
The concept study demonstrated how AI-driven EODT technology could help bring together data from different sources, such as oceans, atmosphere, cryosphere, land, and space weather, as well as deliver useful products to decision makers, the research community, industry, and other public agencies. The concept study also offered recommendations for interoperability and standardization to enhance collaboration with other EODT efforts as part of a next-generation ground enterprise system capable of assimilating observations, operating models, and delivering data products to users.
An EODT is a virtual replica of a real-world environment, system, or object that is continuously updated with data from Earth observation satellites and other sources. This dynamic digital model allows for the simulation, analysis, and prediction of the real-world counterpart’s behavior and performance.
Through JVP, NESDIS collaborates with private industry, academia, and other federal science agencies to explore emerging technologies through concept studies and demonstrations. These efforts help determine the feasibility of new capabilities in meeting NOAA’s mission requirements. While NOAA is not obligated to adopt any technology studied under JVP, the results inform NOAA’s future investment decisions.
Digital twin technology would enable NOAA to simulate past, present, and future environmental conditions to produce predictive models for both forecasting and impact assessments. The AI-driven EODT defined in this study uses machine learning to rapidly integrate data from diverse sources and support models at different scales.
The study made the following recommendations for NESDIS:
- Build on existing digital twin projects currently in development within NOAA, across other federal agencies, academia, and industry by aligning data formats and standards.
- Design a flexible EODT architecture framework that includes a data archive, common data file formatter, data input service, service for containerized algorithm processing, a common processed file and a tiled processed file.
- Improve NOAA-wide data standards to streamline data ingestion.
- Consider a separate digital twin for space weather due to its unique spatial coordinate systems and spatial scales.
- Develop multiple digital twins at various scales (process, domain, and global) with each focused on a particular aspect of the Earth system. The authors believe this will be a more cost-effective solution rather than trying to create a single unified digital twin.
- Apply Observing System Simulation Experiments to determine how best to integrate large-scale or small-scale processes into the digital twin infrastructure.
Industry-led studies such as this, carried out through JVP, provide valuable insights that help inform NOAA NESDIS decisions on future investment decisions for new business models, mission concepts, instruments, spacecraft and other observing systems.