Hydraulic fracture-induced microseismic events in producingoil and gas fields are usually small, and noise levels are high at thesurface as a result of the heavy equipment in use. Similarly, innonhydrocarbon settings, arrays for detecting local earthquakeswill benefit from reduced noise levels and the ability to detectsmaller events will be increased.We propose a frequency-dependentmultichannel Wiener filtering technique with linear constraintsthat uses an adaptive least-squares method to remove coherentnoise in seismic array data. The noise records on severalreference channels are used to predict the noise on a primarychannel and then can be subtracted from the observed data. On atest with an unconstrained version of this filter, maximal noisesuppression leads to signal distortion. Two methods of imposingconstraints then achieve signal preservation. In one casestudy, synthetic signals are added to noise from a pilot deploymentof a hexagonal array nine three-component seismometers,approximately 150150 m above a gas field; noise levels aresuppressed by up to 11 dB at 2–10 Hz. In a second case study,natural seismicity recorded at a dense array 10 m spacing inItaly is used, where the application of the filter improves the signal-to-noise ratio S/N more than 20 dB at 2–15 Hzusing35 stations. In both cases, the performance of the multichannelWiener filters is significantly better than stacking, especiallyat lower frequencies where stacking does not help tosuppress the coherent noise. The unconstrained version of thefilter yields the best improvement in signal-to-noise ratio,but the constrained filter is useful when waveform distortion isunacceptable.
展开▼
