Summary
In this paper we evaluate the use of an automatic moment tensor inversion algorithm on passive seismic data from the Eagle Ford for its usefulness in evaluating the microseismic source mechanisms as well as potential applications for real time processing by comparing it to a hand-picked source mechanism inversion method. Hand picking involves manually picking first arrival p-wave amplitudes for a subset of events and utilizing a grid search for a pure shear or DC (double couple) strike, dip, and rake source mechanism solution (Aki and Richards, 1980). A least squares inversion (Sipkin, 1982) is utilized for the full moment tensor solution. This provides a general fit solution that is representative of all events. The solution fit is verified by applying polarity reversals to correct for the radiation pattern at the surface such that the first p-wave arrival becomes consistent. This is repeated until enough general solutions exist to properly correct move-outs for all events, classifying them into discrete mechanism groups. Automatic picking utilizes a form of full waveform inversion to calculate moment tensors for every microseismic event individually.
The dataset analyzed in this paper exhibited two distinct source mechanism solutions from the hand picking process, one dip-slip style class of events and a strike-slip class extending away from the well as a long coherent trend. This trend was dominated by a strike slip shear mechanism solution causing it to stand out against the rest of the treatment and was interpreted as a sub-seismic fault. Clustering analysis can be used as an early alert system to show the emergence of such a fault or other geo-hazard during hydraulic fracturing operations.
Automatic calculation of moment tensors offers a distinct advantage over the hand picking method in that it enables fast and efficient evaluation of numerous microseismic events and their possible source mechanisms by computer with little need for analyst intervention. While hand picking source mechanism solutions results in a small number of discrete solutions fitted to the entire point set, automatic picking offers a full moment solution for all events. This facilitates a more detailed and accurate picture of stress and resultant fracture network both spatially and temporally. It also offers a much quicker response to large changes in source mechanism in real time allowing the processing to adapt to these changes as they happen. This eliminates the need for an analyst to pull data, pick a new source mechanism, apply it to the processing, and delaying results delivery by requiring reprocessing of data. Such detailed moment tensor analysis makes it possible to quickly build more accurate discrete fracture network models, meaning that engineers can begin to plug microseismic data into their modeling in near real time in a more deterministic way. The most direct application of this type of analysis is real time stress evolution analysis. Being able to determine temporal changes in the stress fields offers a look into how the treatment is changing the stress field as its happening.
