← Back to Sessions

Separating Broad-Band Site Response From Single-Station Seismograms

Description: 

We develop a deep-learning model, SeismAmp, to disentangle site effects in a broad frequency range (0.2 - ~20 Hz) from source and path effects in single-station seismograms (features). Ground-truth data (labels) are homogeneously created using a classical multi-station approach - the generalized spectral inversion at a total number of 1725 sites in the Fourier domain. We tested different feature representations and found that the individual components of each record carry salient information on site response, especially at high frequencies. However, part of the information is lost in its horizontal-to-vertical spectral ratio. The impacts of the number of recordings per site and record selection criteria (by earthquake magnitude, hypocentral distance, focal depth, or at random) are also investigated. When tested at new sites, SeismAmp achieves the lowest standard deviation among tested single-station techniques. SeismAmp could lead to improved site-specific earthquake hazard prediction in cases where site-specific recordings are available or can be collected. It is also a convenient tool to remove repeatable site effects from ground motions, which may benefit other applications, e.g., improving the retrieval of seismic source parameters.

Session: High-frequency Ground Motion Measurements, Assessments and Predictions [Poster]

Type: Poster

Date: 4/18/2023

Presentation Time: 08:00 AM (local time)

Presenting Author: Chuanbin Zhu

Student Presenter: No

Invited Presentation: 

Authors