Using the Reference Noise Algorithm to Suppress Long Duration Noises from Seismic Data

Seismic reflection data are always contaminated by environmental and coherent noises, which some of them have short and others have long duration. Long duration noises have usually a quasi-sinusoidal wave train form. Surface waves are coherent noises, and noises generated by wire agitations, earth agitation due to vehicles traffic, and or noises which resulted from power line are environmental quasi-sinusoidal noises type. For attenuation of these noises there are several suggested methods that each of them has its own advantages and limitations. In this approach, these noises were considered as reference noise, assuming that the noise is sinusoidal wave train with phase and amplitude varying with time. To obtain amplitude of the reference noise, the envelope of absolute value of data is depicted. To calculate its phase, firstly the frequency content of these noises versus time was calculated using a wavelet transform. Then by integrating of calculated frequency, the phase was obtained. In the next step, to construct a trace containing noise, a filter was designed and convolved with reference noise. Then by subtraction of the constructed trace containing noise from input seismic trace, the observed noise was highly attenuated.