PREDICTIVE DECONVOLUTION BASED ON SEISMIC WAVEFORM DIAGNOSTICS FOR ENHANCED MARINE IMAGING RESOLUTION

Abstract

Short-period multiples are a persistent problem in marine seismic processing, particularly in shallow-water environments where near-surface reverberations interfere with primary reflections and decrease temporal resolution. Predictive deconvolution remains a widely used method for attenuating such coherent noise. However, conventional implementations often apply fixed operator parameters, limiting their adaptability to waveform variations across time and offset. This study introduces a predictive deconvolution framework guided by seismic waveform diagnostics, in which operator parameters specifically prediction lag and filter length are selected based on trace characteristics such as waveform periodicity and spectral energy distribution. The approach is designed to improve multiple suppression while preserving the fidelity of primary reflections on a 2D marine pre-stack seismic dataset acquired in a shallow offshore setting characterized by strong short-period multiples and limited bandwidth. The results demonstrate around 25% increase in frequency bandwidth, improved reflector continuity, and reduced coherent noise in pre-stack gathers. Compared to conventional deconvolution, the waveform informed design achieves a more effective balance between attenuation and resolution. The proposed approach is applicable to modern marine datasets where high-resolution imaging is limited by near-surface interference.