Recommendations from error analysis of single-well microseismic data with full-wavefield moment tensor inversion: A case study

Microseismic events and their source mechanisms play a significant role in our understanding of hydraulic fracturing. To better identify the reliability of source mechanisms, we examine the limitations of microseismic field data imposed by (1) lack of angular coverage, (2) moment tensor inversion constraints, and (3) effects of mild anisotropy. We use synthetic seismograms to examine cases of either pure double-couple (DC) or compensated-linear-vector dipole (CLVD) sources. The open-source full-wavefield moment tensor inversion code (ISOLA) incorporates both near- and intermediate-field terms, which can increase the accuracy of the inversion if source-receiver distances are small. The tested locations and dominant source frequencies of the synthetic seismograms used for analysis resemble the expected locations and dominant frequencies of microseismic events extracted from a multi-stage field data set in the Barnett Shale of East Texas. We conclude that although a horizontal receiver array can provide greater angular coverage of vertical failure planes than a vertical receiver array, the strike of shear (DC) sources cannot be accurately resolved unless the receiver array has angular coverage to sample both sides of the shear failure plane. If the source is DC, the inversion can result in a CLVD mechanism that is overestimated by up to ~40% and if the source is CLVD, the DC mechanism can be overestimated by ~20%. Errors in the inversion results are interpreted to be because of the lack of receiver angular coverage of the source rather than possible errors associated with source mislocation. The use of the deviatoric assumption decreases the error in the resolved source mechanism by ~30-~40%, however, this increases the error in the resolved source strike to ~15�-~35�. For pure shear sources, 5% VTI anisotropy in the medium has minimal effect on source orientation (<15�) but can introduce 25-50 m of error in the source location. The neglect of anisotropy in moment tensor inversion has a greater effect (~40% variation) on the estimated source mechanism for pure CLVD sources than for pure DC sources.