Signal Extraction and Automated Polarization Analysis of Multicomponent Array Data

We present a method to extract seismic signals from three-component array data and estimate their polarization properties at each station. The technique is based on a singular value decomposition (SVD) of the complex three-component analytic signal and applies to linearly as well as elliptically polarized seismic phases. To increase accuracy we simultaneously analyze data from different stations and apply a noise weighting based on prearrival data. For polarization analysis, an au- tomated routine is also included. The automated routine selects the data window with the best signal-to-noise ratio from which to obtain a polarization. A linearity measure and a confidence interval accompany the polarization estimate at each station in the array. We test our technique for automated polarization analysis on synthetic P-wave data and compare results with those from other methods. A microseismic dataset from the North Sea provides a unique opportunity to statistically compare previous and independently obtained P-wave polarizations with those provided by the auto- mated technique presented here. We conclude that, for P-wave polarization analysis, our method is robust and significantly more accurate than conventional, mainly man- ual methods. This is especially so on data with polarized and correlating background noise. It is also faster and provides meaningful quality estimates.