Fault mechanisms from near-source data: joint inversion of S polarizations and P polarities

We propose a non-linear inversion method for studying the earthquake mechanism by combining the information carried by both S-wave polarizations and P-wave polarities from near-source records. The posterior probability of parameters (strike, dip and slip fault angles) for the given observational data sets is computed by using a Bayesian approach. The conditional probability density function of S polarizations given a model parameter set is defined assuming a Gaussian distribution for the expected errors. The P polarity information is taken into account in the form of a prior probability density function, which has been defined according to Brillinger, Udias & Bolt (1980). The method is based on the estimate by an exhaustive search of the posterior probability of model parameters. This probability is then represented by its projection on plane sections in the model space. This enables one to locate the maximum likelihood solutions, and to get a reliable estimate of the parameter correlation and resolution. Numerical examples and data analysis show that the addition of a few S polarizations to a P polarity data set greatly improves the resolution on the fault parameters. The proposed method is suitable to study the low-magnitude seismicity in active tectonic and volcanic areas which are monitored by local networks as the use of three-component sensors becomes more systematic.