Moment tensor migration imaging

We develop and apply an imaging procedure for simultaneous location and characterization of seismic source properties called Moment Tensor Migration Imaging. The procedure constructs images for moment tensor components using a weighted diffraction stack migration, and combines ray-theoretical Green's functions with a reverse time moment tensor imaging methodology. By applying an approximation we term the ray-angles only approximation', we form an expression for Moment Tensor Migration Imaging where the migration weights depend only on the take-off and arrival angles for rays leaving receiver positions and incident upon the image points. Moment Tensor Migration Imaging retains the benefits of diffraction stack procedures for source location and characterization, namely speed, flexibility, and the potential for incorporating non-linear stacking procedures, whilst also providing the benefits of moment tensor imaging such as: the inclusion of multiple phase and multiple component data; the collapsing of the source radiation pattern; estimation of the moment tensor. We examine variations of the imaging procedure through a synthetic test. We show that although the assumptions required for the imaging and ray-angles only approximation may not be strictly valid for realistic survey geometries, a simple weight adjustment can be used to obtain more accurate and stable results in these situations. In our synthetic example we find that the use of a P-wave only migration without this reweighting structure produces poor results, whereby the resulting images show activity upon incorrect moment tensor components. However, many of these effects are mitigated by use of the reweighting scheme and the results are further improved through the introduction of non-linear stacking operators such as semblance weighted stacks. The highest quality moment tensor images (for the synthetic test examined here) are obtained through the use of both P-wave and S-wave wave fields. This highlights the importance of multicomponent data and multiphase modelling when characterizing seismic sources. We also find that the imaged moment tensor components vary proportionately when the input velocities are perturbed by a scale factor. This suggests, for the geometry investigated here, derived source properties such as fault-plane solutions and shear-tensile components will not be influenced by bulk changes in seismic velocities. Finally, we show the application to a real microseismic event observed using a surface array during hydraulic fracturing. We find that the procedure collapses the seismic radiation pattern into an anomaly with a maximum at the hypocentre and our derived mechanism is consistent with the observed radiation pattern from the source.