S046-0022
The Finite-Fault Rupture Detector (FinDer): Recent Performance and Developments
The Finite-Fault Rupture Detector (FinDer): Recent Performance and Developments
Monday, 14 December 2020
Poster
Abstract:
Rapid information on fault rupture extent is important for estimating seismic ground-motions and damage in large earthquakes as needed for earthquake early warning and rapid response. The Finite-Fault Rupture Detector (FinDer) algorithm models a line-source in real-time by comparing seismic amplitudes with pre-computed templates (Böse et al., 2018). Here we summarize the recent performance and developments of FinDer.
FinDer is one of the algorithms adopted by the US West Coast ShakeAlert warning system. The largest real-time detected earthquake was the 2019 M7.1 Ridgecrest earthquake. While the performance of ShakeAlert was generally good, FinDer and EPIC both underestimated the magnitude. In the case of FinDer, this resulted from increasing data latencies in the network and improper handling of abnormally late data from closeby stations. We fixed this in the latest FinDer release and developed a new method to improve magnitude convergence.
In a retrospective study, we simulated the performance of FinDer for three large earthquakes in Sichuan, China, including the 2008 M7.9 Wenchuan earthquake. We found that, if the seismic data recorded by the China Strong Motion Network had been available in real-time, 50-80% of sites experiencing shaking with MMI IV-VII and 30% experiencing MMI VIII-IX could have been issued a warning with lead times of 10s and 5s, respectively.
To address the problem that most countries do not operate dense sensor networks, we have explored how FinDer would perform on crowd-sourced felt-intensity reports collected by the EMSC for large earthquakes. The FinDer models achieve a good agreement with published finite-source models: for 50% the difference in strike is less than 30 deg, and for 75% less than 60 deg. FinDer models based on felt-reports could be computed automatically within 10 to 30 min after large global earthquakes – to provide important information on expected losses and support rescue efforts.
The next generation of FinDer will include slip distribution estimates and moment magnitudes derived from the back-projection of dynamic displacement amplitudes onto the FinDer line-source (FinDer-S), and the prediction of future rupture evolution based on gradients of material properties associated with along-strike variations of fault maturity (FinDer-S+).
FinDer is one of the algorithms adopted by the US West Coast ShakeAlert warning system. The largest real-time detected earthquake was the 2019 M7.1 Ridgecrest earthquake. While the performance of ShakeAlert was generally good, FinDer and EPIC both underestimated the magnitude. In the case of FinDer, this resulted from increasing data latencies in the network and improper handling of abnormally late data from closeby stations. We fixed this in the latest FinDer release and developed a new method to improve magnitude convergence.
In a retrospective study, we simulated the performance of FinDer for three large earthquakes in Sichuan, China, including the 2008 M7.9 Wenchuan earthquake. We found that, if the seismic data recorded by the China Strong Motion Network had been available in real-time, 50-80% of sites experiencing shaking with MMI IV-VII and 30% experiencing MMI VIII-IX could have been issued a warning with lead times of 10s and 5s, respectively.
To address the problem that most countries do not operate dense sensor networks, we have explored how FinDer would perform on crowd-sourced felt-intensity reports collected by the EMSC for large earthquakes. The FinDer models achieve a good agreement with published finite-source models: for 50% the difference in strike is less than 30 deg, and for 75% less than 60 deg. FinDer models based on felt-reports could be computed automatically within 10 to 30 min after large global earthquakes – to provide important information on expected losses and support rescue efforts.
The next generation of FinDer will include slip distribution estimates and moment magnitudes derived from the back-projection of dynamic displacement amplitudes onto the FinDer line-source (FinDer-S), and the prediction of future rupture evolution based on gradients of material properties associated with along-strike variations of fault maturity (FinDer-S+).