← Back to Sessions

Kinematic Source Variability in Ground-motion Simulations and Implications for Seismic Hazard Analysis

Description: 

Traditional Probabilistic seismic hazard assesment (PSHA) uses empirical ground-motion models (GMMs) under the ergodic assumption, often overestimating hazard for low-probability events. Simulated ground-motions eliminates the ergodic assumption as they are explicitly site-based and the variability is from variations in source parameters and wave-propagation effects. While physics-based earthquake rupture source characterizations enable efficient rupture scenario generation, understanding the contribution of source parameters to total variability is crucial for their application. In this study, we use a new rupture generator based on a machine-learning method applied to previous rough-fault dynamic rupture simulations to generate a suite of rupture scenarios for a Mw 6.5 strike-slip event. In total, we generate 1080 rupture scenarios with variations in fault length (5 variations), slip realizations (8), hypocenter locations (9) and average rupture velocity (3). We compute broadband seismograms within a 100 x 100 km computational domain in a 1D crustal velocity model at 441 stations (Rjb < 65km) upto a maximum frequency of 7 Hz.

Our results show that while the between-event variability (τ) of the simulations aligns with GMMs, the within-event variability (ϕ) exceeds GMM estimations. Using mixed-effects models, we partition variability across input parameters and azimuths. We observe that in regions of directivity (±45° from fault strike), variations in hypocenter position along strike account for ~90% of τ, independent of spectral period. Conversely, in fault-perpendicular regions, slip and fault length explain ~80% of τ. Hazard computations in terms of probabilities of exceedance reveal under- or overestimations relative to GMMs when using a single fault geometry. Finally, by incorporating fault-length and hypocenter location into a logic tree framework with weighting scheme from the respective empirical models, we compute mean exceedance curves. This study underscores the importance of source parameter variability in reducing bias and improving the reliability of PSHA using physics-based simulations.

Session: Recent Advances in Modeling Near-source Ground Motions for Seismic Hazard Applications [Poster]

Type: Poster

Date: 4/16/2025

Presentation Time: 08:00 AM (local time)

Presenting Author: Paul Martin

Student Presenter: Yes

Invited Presentation: 

Poster Number: 51

Authors