Moment Tensor Inversion of Microseismic Events Using Strain Waveforms Recorded by Multi-Well-Distributed Acoustic Sensing

ABSTRACT Distributed acoustic sensing (DAS) array seismic data are now routinely acquired on fiber-optic cables in wells for downhole monitoring of microseismic events during geothermal and oil and gas operations. In an enhanced geothermal system experiment in Blue Mountain, Nevada, downhole DAS arrays operational in three different wells at all times provided unprecedented constraints on the locations and source radiation patterns of the microseismic events. We develop a simple framework for inverting low-frequency (∼4 to 16 Hz) complete axial strain waveforms recorded by DAS for point-source moment tensors using axial strain Green’s functions calculated for a 1D velocity model using frequency–wavenumber integration. We fit the strain waveforms recorded at two to three wells with a variance reduction of ∼36% to 74%. We constrain the best-fitting source types to be strongly deviatoric; assuming a double-couple mechanism, we obtain predominantly normal faulting and right-lateral strike-slip faulting mechanisms on northwest to north-northwest-trending faults that agree with the geologic knowledge of the study area and the extensional Basin and Range tectonics prevalent in Nevada. We further validate magnitudes derived from low-frequency strain spectral amplitudes of far-field body waves by comparing them with magnitudes derived from moment tensors. Finally, for two of the larger events, we find the strike-slip focal mechanisms derived from the DAS data to be consistent with those estimated from low-frequency, vertical-component displacement waveforms recorded by a sparse network of surface seismometers. Moment tensor inversion applied to downhole DAS arrays helps characterize the size of the microseismic events that are relevant to induced seismicity, and the state of stress in the subsurface during geothermal, fracturing, and oil and gas operations. The inversion framework is general, and with a sufficiently accurate velocity model and good signal-to-noise ratio at frequencies lower than the corner frequency, it can be extended to dark fiber DAS arrays at the surface as well.