Collaborative research: frameworks: geophysical AI-driven integration & assimilation (gaia)

The National Science Foundation Office of Advanced Cyberinfrastructure awarded Earthscope Consortium Inc. (operating as Iris) $224,724 on August 1, 2026, under the Computer and Information Science and Engineering program (CFDA 47.070) to develop and deploy open-source cyberinfrastructure that integrates geophysical sensor data, satellite imagery, meteorological observations, and groundwater measurements with numerical simulations and artificial intelligence to create continuously updating digital twins for real-time hazard estimation of earthquakes, landslides, and floods. The project builds unified computing tools that span from laptops to commercial cloud platforms and national high-performance computing centers, linking physical simulations across distributed systems. The research team conducts five connected lines of work, with field testing on landslides in Alaska and the Pacific Northwest. The project also develops tutorials, online workshops, and graduate-level coursework to lower adoption barriers for computational science and trains students in the use of these tools. The infrastructure framework is designed as a template for data-rich fields beyond geophysics. The award period runs from August 1, 2026, through July 31, 2032, with place of performance in Washington, DC.

Abstract

Earthquakes, landslides, and floods cause hundreds of billions of dollars in damage and take thousands of lives each year. These hazards are often linked: heavy rain can saturate a hillslope, an earthquake can weaken that same slope, and the slope may then fail during a later storm. Forecasting where and when such events will strike remains hard, in part because the measurements that could warn of them, such as ground vibrations and deformation, satellite imagery, meteorological data, and groundwater levels, are gathered by separate systems that rarely work together. This project builds open shared computing tools that combine these streams of data with numerical simulations of the ground and water dynamics, producing a continuously updating digital twin of a region that can estimate hazards in close to real time. The investigators test the full system on landslides in alaska and the pacific northwest. By making advanced computation and artificial intelligence (AI) simpler to use, the project lowers the adoption barriers for this kind of research for all scientists. The project also trains students through open tutorials, online workshops, and graduate courses. The same tools, built here for geophysical hazards, are designed to serve as a template for other fields that are rich in data and physical models. By helping protect communities from natural disasters and by broadening who can take part in computational science, the project advances scientific knowledge and the national welfare. The geophysical AI-driven integration and assimilation (gaia) project develops a unified cyberinfrastructure that links physical simulations, sensor observations, and artificial intelligence across computing systems ranging from laptops to commercial cloud platforms and national high-performance computing centers. A team of researchers carries out five connected lines of work. First, the team is assembling curated, analysis-ready datasets that pair records of past landslides and ground failures with seismic, geodetic, and hydrologic sensor networks, with weather observations, and with satellite imagery that preceded and accompanied the landslides and ground failures. These observations are combined with matching physics simulations and a corpus of scientific literature to build a curated knowledge base. Second, the team builds a data aggregator, a unified software client that gathers data from many national archives and returns IT in consistent, cloud-native formats ready for modeling across computing systems. Third, the team designs methods that fuse observations of differing spatial and temporal resolution, combining, for example, frequent measurements at fixed points with detailed but infrequent satellite maps. Fourth, the team trains fast surrogate models that approximate expensive physics simulations and couples them to incoming observations through data assimilation to produce real-time digital twins that forecast geohazards. Fifth, the team prototypes research-assisting software agents that retrieve and reason over selected literature and code to recommend methods, flag knowledge gaps, and generate reproducible workflows. The resulting open software, datasets, and trained models give the research community scalable methods for combining diverse data with physical models, an approach useful well beyond hazard science. This award by the office of advanced cyberinfrastructure is jointly supported by the research, innovation, synergies and education section in the directorate of geosciences. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the foundation's intellectual merit and broader impacts review criteria.- subawards are not planned for this award.