Application of frequency-dependent multichannel Wiener filters to detect events in 2D three-component seismometer arrays

Abstract

Hydraulic-fracture induced microseismic events are usually small, and noise levels are high at the surface due to the activities associated with a producing oil field. Similarly, local arrays for the detection of local earthquakes will also benefit from reduced noise levels and detect smaller events. We present a frequency-dependent multi-channel Wiener filtering technique with linear constraints, which employs an adaptive least-squares technique to remove coherent noise in seismic array data. The noise records on a number of reference channels is used to predict the noise on a primary channel, which can then be subtracted. We implement and test first an unconstrained version of this filter, where maximal noise suppression can lead to signal distortion. Two methods of imposing constraints are then introduced to achieve signal preservation. We test this technique with two case studies. First, synthetic signals are added to actual noise from a pilot deployment of a hexagonal array (9 three-component seismometers, approximate size 150 m × 150 m) in an oil field; noise levels are suppressed by up to 11 dB (at 1 - 6 Hz). Secondly we use natural seismicity recorded at a dense array (∼10 m spacing) in Italy where the application of the filter reduces the signal-to-noise ratio by more than 20 dB (at 8 - 15 Hz), using 35 stations. In both cases, the performance of the multi-channel Wiener filters is significantly better than stacking, especially at lower frequency where stacking does not help to suppress the coherent noise. The unconstrained version of the filter yielded the best improvement in the signal-to-noise ratio, but the constrained filter is useful when waveform distortion is not acceptable.