Automated detection and location of seismic events on Piton de la Fournaise volcano by waveform migration

We present the continued development of WaveLoc, and automated seismic event detection and location algorithm based on waveform migration, that therefore bypasses the phase-picking and association phases common to most automated location algorithms. WaveLoc is a 3-step process : 1) we filter and calculate the kurtosis of the raw waveforms in order to highlight the non-stationary characteristics of seismic events; 2) we migrate and stack the first derivatives of the kurtosis waveforms, which highlight the P-wave arrivals, according to an a-priori P-wave velocity model (1D or 3D); 3) we detect and simultaneously locate seismic events by analyzing the local maxima of the resulting 3-D time-dependent stacks. We have applied the WaveLoc algorithm to the seismic swarms recorded on the Piton de la Fournaise volcano (Reunion Island) between 2009 and 2011, using data from the UnderVolc experiment (Brenguier et al., 2012) and the Prono et al. (2009) 3D velocity model. We compare the locations obtained using WaveLoc to those obtained from manual picking in order to evaluate the robustness of the automated algorithm. Automated location of single events is in general limited by "picking" errors (in our case intrinsic variations in stationarity properties of the seismic signals) and inadequacies in the a-priori velocity models. In order to improve both accuracy and precision of our locations, we have systematically searched for multiplets by cross-correlation, and relocated these multiplets using a simple double-difference algorithm.