Multievent Moment-Tensor Inversion for Ill-Conditioned Geometries

Xin Yu,Scott Leaney,Jim Rutledge,Chris Chapman

DOI: https://doi.org/10.1190/geo-2015-0074.1

IF: 3.264

2016-01-01

Geophysics

Abstract:Moment-tensor inversion under single monitoring well geometries becomes unstable due to the singularity of the inversion matrix. But microseismic events observed during hydraulic fracturing commonly show clusters of events with similar source mechanisms despite differences in the origin time and the magnitude. If the events with similar source mechanisms can be grouped and inverted for a single common moment tensor, the singularity can be eliminated. We have developed a normalized multievent moment-tensor inversion (NME-MTI) method, which does the MTI simultaneously for multiple events, to test the feasibility of this multievent approach. First, the scalar moment for each event was estimated using the far-field low-frequency level at each receiver. Then, the displacements measured at the receivers were normalized by the scalar moment and used to invert for the common moment tensor simultaneously for all the events in the group. We introduced a gradient search method to minimize the overall misfit by adjusting the scalar moment for each event to reduce the errors introduced by the scalar moment estimation. The algorithm was tested with a synthetic data set with four monitoring wells and a field data set with dual monitoring wells. It was proved that the NME-MTI method can retrieve the moment tensors of the event groups with data from a single monitoring well. The effects of uncertainties on the inversion were examined with data noise, scalar moment uncertainty, and event location uncertainty. The results showed that the ME-MTI result is much less sensitive to the data noise and the scalar moment uncertainty than the single-event approach. The results also determined that although the bias to the solutions increases when the event location uncertainty increases, the bias can be controlled by reducing the event location uncertainties using a more accurate location algorithm.