Bayesian inversion with Markov chains—II. The one-dimensional DC multilayer case

In this paper, we will report on the application of Bayesian inference to DC resistivity inversion for 1-D multilayer models. The posterior probability distribution is explored through a Markov process based upon a Gibbs's sampler. The process would lead to unrealistic estimates without additional prior information, which takes the form of a second Markov chain where the transition kernel corresponds to a smoothness constraint. The outcomes are posterior marginal probabilites for each parameter, as well as, if required, joint probabilities for pairs of parameters. We will discuss the main properties of the method in the light of a theoretical example and illustrate its capabilities with some field examples taken from various contexts.     

Linear programming approach to moment tensor inversion of earthquake sources and some tests on the three-dimensional structure of the upper mantle

Summary. A new method of moment tensor inversion is developed, which combines surface wave data and P -wave first motion data in a linear programming approach. Once surface wave spectra and first motion data are given, the method automatically obtains the solution that satisfies first motion data and minimizes the L1 norm of the surface wave spectra. We show the results of eight events in which the method works and is stable even for shallow events. We also show one event in which surface wave data and P -wave first motion data seem to be incompatible. In such cases, our method does not converge or converges to a solution which has a large minor (second) double couple component. It is an advantage that the method can determine the compatibility of two data sets without trial and error.
Laterally heterogeneous phase velocity corrections are used to obtain spectra at the source. The method is also applied to invert moment tensors of eight events in two recent three-dimensional (3-D) upper mantle structures. In both 3-D models, variances of spectra are smaller than those in a laterally homogeneous model at 256 s. Statistical tests show that those reductions are significant at a high confidence level for five events out of eight examined. For three events, we examined those reductions at shorter periods, 197 and 151 s. The reduction of variances is comparable to the results at 256 s and is again statistically significant at a high confidence level. Orientation of fault planes does not change very much by incorporation of lateral variations of phase velocity or by doing inversions at different periods. This is mainly because of the constraints from P -wave first motion data. Scatter of phase spectra at shorter periods, especially at 151 s, is great and suggests that surface wave ray paths deviate from great circle paths substantially and these effects cannot be ignored.