Marine Dispersive Noise Removal Using f-P Modelling

In shallow marine seismic data where a hard water bottom is present, post-critical multiple energy can become trapped in the water layer creating dispersive noise modes. The phase velocities of this noise are sensitive to the near surface parameters which can vary significantly across a survey. In certain geological settings, some of these modes dominate the shot record but cannot be attenuated by traditional demultiple techniques such as Tau-P deconvolution as their period is too short. We present a simple method of attenuating these modes by modelling them in the frequency/slowness (f-P) domain. After Tau-P deconvolution, the slowness and dominant frequency of the strongest remnant noise mode is automatically picked for each shot record in a survey. This allows the mode to be isolated, modelled and subtracted from the shot record. Automated f-P modelling was tested on a large marine 3D survey over the North-West Shelf, offshore Australia where shallow carbonates cause complex dispersive noise. This data driven technique was shown to reliably model this dispersive noise, allowing it to be attenuated and reveal underlying primary energy.