← Back to Sessions

Enhancing Earthquake Early Warning With Real-time Ground Motion Assimilation for Rupture Directivity Effects via Kalman Filter

Description: 

Earthquake early warning (EEW) is the primary approach for mitigating earthquake hazards, offering the public a lead time of a few to tens of seconds before destructive ground shaking arrives. The widely used point-source algorithm has achieved great success in estimating timely and reasonable ground motion values for given sites, provided rapidly-determined source magnitude and location. However, its accuracy is limited by several factors. First, rupture directivity, a critical source property that amplifies shaking amplitude in specific directions, is not accounted for in most EEW frameworks. Second, Ground Motion models (GMM) used in EEW do not capture event-to-event variability, leading to missed or false alerts. Furthermore, in real-world scenarios, underestimates of earthquake magnitude derived from early P-wave data often occur due to instrumental saturation issue and long rupture process of large events also increase these uncertainties and lead to inaccurate warnings. To address these challenges, we apply the Kalman Filter method to calibrate and update GMM coefficients using real-time peak amplitude data from stations, accounting for variability across different events. By including a new rupture directivity term, our anisotropic GMM also captures directional ground shaking amplification in real time. Applications to several M6+ earthquakes in Taiwan demonstrate the efficacy and relevance of our approach for EEW systems. More importantly, our method can handle the issue of magnitude underestimation during the EEW. For the April 3, 2024 Mw 7.4 Hualien earthquake that caused missed warnings in seven prefectures (including the Taipei metropolis) due to lower estimate of the initial magnitudes, the method can rapidly calibrate predicted ground motions and identify primary rupture direction within 20-25 seconds after origin time, successfully addressing the missed alerts during this event. The data-assimilated anisotropic GMM via the Kalman Filter could provide a practical solution for accurate ground motion predictions and timely source property information, supporting post-event emergency response efforts.

Session: Performance and Progress of Earthquake Early Warning Systems Around the World - II

Type: Oral

Date: 4/16/2025

Presentation Time: 11:15 AM (local time)

Presenting Author: Yi-Sheng

Student Presenter: No

Invited Presentation: 

Poster Number:

Authors