Earthquakes generate ground shaking that can be dangerous to people and infrastructure. To better prepare for future earthquakes, scientists often simulate how the ground might move during different scenarios. However, this is hard to do accurately, especially when there isn’t much information about the earthquake source or local geology.

We developed a new system called HEGGS that uses artificial intelligence to create realistic earthquake shaking waveforms. Uniquely, HEGGS only needs four basic pieces of information to work: where the earthquake occurred, how deep it was, how strong it was, and where the shaking is being recorded. No detailed geological data is required.

We trained and validated HEGGS using real earthquake data from North America, East Asia, and Europe. It can simulate ground shaking that closely matches real seismic recordings, including key features like the arrival of P- and S-waves and how strong the shaking is at different locations. It also works for new scenarios, such as hypothetical large earthquakes or virtual seismic stations.

HEGGS can help researchers, engineers, and emergency planners explore 'what-if' earthquake scenarios, especially in areas where detailed local data is lacking. This makes it a valuable tool for improving earthquake preparedness and hazard assessments around the world.