A matched filter technique that uses cross-correlation and migration of the recorded waveforms has been successfully used to relatively locate microseismic events. In addition to producing consistent relative locations, the matched filter corrects for radiation pattern effects and near-surface structure. The relative locations produced using this methodology were compared with a proprietary, direct location method. The new results produce a solution set that reveals two parallel trends of microseismic events which are interpreted as 1500' long fracture zones approximately 100' wide. We observed asymmetric fracture growth and re-fracturing of the previously stimulated zones. Time correspondence of the observed evolution of seismic events with engineering pump curve data reveals approximate linear growth rates of several feet per minute and possible proppant placement along the induced fractures.
Distinguishing signal from noise is an important goal of many scientific endeavors. It is essential to the seismic industry. Here, we apply the matched filter technique of Eisner et al. (2008) to enhance signal from induced microseismic events detected on a surface seismic array. This technique was originally developed for detection of weak signals from distant nuclear explosions (e.g. Gibbons and Ringdal, 2006) and the study of very small earthquakes (so called 'nanoseismology' - e.g. Joswig, 2008).
The matched filter approach, as used in this study, is a cross-correlation of waveforms from good signal-to-noise ratio events ("masters") with noisier events ("slaves") on corresponding receiver elements. This filter enhances consistent stacking of P-wave energy as it corrects for local variations of near surface structure. In addition, it corrects for polarization changes due to complex source radiation effects because phase polarities correlate positively. The filtered data are then combined with a migration based stacking technique to locate microseismic events relative to the location of the master event. The absolute location accuracy of the master event is controlled by using only the best signal-to-noise ratio event(s) in the global data set. The relative location accuracy is achieved by migration of cross-correlations of the weaker events correlated with a strong signal-to-noise master.