Deep learning-based Vz-noise attenuation for OBS data

Abstract

Translate Article

In multicomponent ocean bottom seismic (OBS) acquisition, hydrophones and geophones (or accelerometers) are employed to record seismic signals. Combining records from the hydrophone (P-component) and the vertical geophone (Z-component) allows for elimination of the receiver-side ghost and water-layer multiples, when performing wavefield separation above and below the seafloor respectively. Apart from the environmental noise that affects both hydrophone and geophone records, the Zcomponent is often contaminated by an additional type of shear-like noise known as Vz noise. Unless removed, Vz noise can cause significant issues with subsequent processing steps. There have been several works on attenuating Vz noise with different levels of success. Here we propose an effective deep learning (DL) approach for this task.