FMNEAR: Determination of Focal Mechanism and First Estimate of Rupture Directivity Using Near-Source Records and a Linear Distribution of Point Sources

The FMNEAR method is based on both the waveform inversion of near-source seismic records and on a linear finite-source model. The primary source parameters that are determined are the moment magnitude (M-w), the double-couple focal mechanism (strike, dip, and rake), and the distribution of seismic moment along strike, which provides a first-order estimate of rupture length and directivity. Source depth is also explored. An advantage of the approach is the use of the finite-source model directly within the search of the focal mechanism. This strategy allows the inclusion of strong-motion records at short distances for moderate to very large earthquakes (M-w > 5.5). However, for moderate to small earthquakes (M-w < 5.5), the source is simplified to a single point source. A specific band-pass filter is automatically adapted for each individual component of the seismograms (north, east, vertical). The linear finite source is represented by point sources aligned along strike, each point source being characterized by a local source time function (STF). The inversion is carried out in successive steps combining fast grid searches on the (strike, dip, and rake) parameters and simulated annealing to determine rupture onset times and the shape of the local STFs. The nonlinear grid search offers the possibility of assessing the degree of uniqueness of the solution, and a confidence index is defined. The method was tested on 11 earthquakes worldwide, ranging from M-w 4.5 to 9.0. Test cases incorporate continental events recorded by a large number of well-distributed stations and less favorable cases. In its present form, the FMNEAR approach is very well adapted to continental earthquakes surrounded by seismic stations, but it can also give informative results in more difficult configurations. The FMNEAR method is now ready for fully automated determinations and is already implemented in near real time.