Building a digital twin of the ground over six years — joining seismic, deformation, satellite, weather and groundwater data that have never worked together

1Measured separately, and so never joined

Earthquakes, landslides and floods often come linked. Heavy rain saturates a hillslope, an earthquake weakens the same slope, and the slope fails in a later storm. Yet the measurements that could signal any of this — ground vibration, crustal deformation, satellite imagery, weather, groundwater level — are each gathered by a separate system.

The premise of this award is that the difficulty lies not in a shortage of observation but in how separately it is collected.

2A six-year period

Six years is the longest duration among the records this site holds — twice the median of three. The length reflects that what is being built is not a single result but computing infrastructure other researchers will keep using. Infrastructure has to be designed with the time it will be used and maintained included.

3Five lines of work stacked up

The project runs five connected lines. First, assembling curated, analysis-ready datasets pairing records of past landslides and ground failures with seismic, geodetic and hydrologic sensor networks, weather observations and the satellite imagery that preceded and accompanied them. Second, building a data aggregator that gathers from many national archives and returns consistent, cloud-native formats.

Third, designing methods that fuse observations of differing spatial and temporal resolution — frequent measurements at fixed points alongside detailed but infrequent satellite maps. Fourth, training fast surrogate models that approximate expensive physics simulations and coupling them to incoming observations through data assimilation to produce real-time digital twins.

Fifth, prototyping research-assisting software agents that retrieve and reason over selected literature and code to recommend methods, flag knowledge gaps and generate reproducible workflows.

4As a template for other fields

What stands out in the description is the explicit statement that the tools built here are not confined to geophysical hazard forecasting. Any field rich in data and physical models can use the same pattern. Lowering the barrier to advanced computation and AI so more scientists can take part is also named as an aim, alongside training students through open tutorials, online workshops and graduate courses.

Of the 70 distinct programs among the NSF awards this site holds as of 2026-09-02, this award belongs to both AI and Geosciences and Software Institutes. Amounts are the obligated amount as of the check date and may change.