← Back to Sessions

Testing Data-driven Techniques for Separating Free-field Ground-motion From Building Response in Smartphone Data

Description: 

The field of ground-motion modeling has overwhelmingly embraced non-ergodic approaches that explicitly model the large spatial variability in source, site, and path effects in observed ground motions. The field is also increasingly focusing on modeling the Fourier Amplitude Spectrum (FAS), as opposed to peak amplitudes, as a metric that more accurately conforms to modeled physics. While massively improved seismic network coverage has in large part enabled this revolution, the ultra dense resolution sought in urban environments has not yet been attained with traditional sensor networks. A distributed network of smartphones may contribute the high density needed. The MyShake app, a free smartphone app delivering earthquake early warning alerts to users on the US West Coast, offers the citizen science capabilities to harvest this data. However, smartphone data is not free-field, but rather “contaminated” by building response, as previous work has shown. In addition, exact building responses are not known for these massive sensor networks, precluding direct separation of free-field ground motion. In this work, we use observed earthquake waveform data recorded by smartphones in California running MyShake. We use a derived FAS database to test various techniques to separate frequency ranges that capture pure free-field ground motion from those contaminated by building response in the 0.5-25 Hz band. We test an empirical, heuristic-driven technique as well as a method based on machine learning-driven signal reconstruction. Both methods fuse smartphone observations with free-field ground motion observations and non-ergodic model predictions. We test the separation ability of both methods and investigate how smartphone-derived FAS “corrections” can be fused with existing, traditionally derived-non-ergodic effects.

Session: Earthquake Ground Motions and Structural Response: Emerging Tools and Applications - I

Type: Oral

Date: 4/16/2026

Presentation Time: 08:15 AM (local time)

Presenting Author: Savvas Marcou

Student Presenter: Yes

Invited Presentation: 

Poster Number:

Authors