The page content begins here

FinDer (Finite-Fault Rupture Detector) Algorithm

Seismic-ground motions in large earthquakes are controlled by the distance to the rupturing fault, not by the hypocentral distance as is assumed in traditional point-source algorithms. The Finite-Fault Rupture Detector (FinDer) algorithm (Böse et al., 2012) determines fast and robust line-source models of large earthquakes in order to enhance ground-motion predictions for earthquake early warning (EEW) and rapid response. The algorithm quantifies model uncertainties in terms of likelihood functions (Böse et al., 2015), and can be applied across the entire magnitude range from M2 to M9 (Böse et al., 2018).

The FinDer algorithm is based on template matching, in which the spatial distribution of high-frequency ground-motion amplitudes (usually peak ground acceleration (PGA)) in a seismic monitoring network is continuously compared with theoretical template maps. These templates are pre-computed from empirical ground-motion prediction equations (GMPEs) for line-sources of different lengths and magnitudes, and can be rotated on-the-fly to constrain the strike of the earthquake fault rupture.

The template that shows the highest correlation with the observed ground-motion pattern is determined from a combined grid-search and divide-and-conquer approach (Böse et al., 2018). The resulting finite-source model is characterized by the line-source centroid, length, strike and corresponding likelihood functions. The model is updated every second until peak shaking is reached, thus allowing to keep track of fault ruptures while they are still evolving.

Compared to other, more traditional EEW algorithms, FinDer has a number of interesting features (see Böse et al. (2018) for details):

  • Characterization of seismic ground-motions rather than earthquake sources;
  • Consistent models and uncertainties for both small and large earthquakes;
  • No magnitude saturation in large earthquakes;
  • Applicable to complex earthquake sequences;
  • No station averages, but true network solutions;
  • Independent from traditional phase picker and pick associators;
  • Unlikely to trigger during teleseisms;
  • Enables joint seismic-geodetic real-time finite-fault models (e.g. for tsunami warning);
  • Can resolve fault-plane ambiguities, including those of small earthquakes.

FinDer v.2 is coded in C++ and embedded in both SeisComP3 and Earthworm/AQMS, which allows us to apply and test FinDer in various seismic real-time networks around the globe and thus in various tectonic settings. At the moment, we are running FinDer in real-time in the ShakeAlert Production system, which will be rolled-out to the general public in California, Washington, and Oregon in 2018. We are also testing FinDer in real-time in Switzerland, Chile, El Salvador, Costa Rica, and Nicaragua. FinDer is enhanced and further developed in various research projects, including SERA and FAULTS_R_GEMS.

The implementation of FinDer proceeds in close collaboration with the US Geological Survey (USGS) and the California Institute of Technology (Caltech).