Multiples Suppression in Common-Offset GPR Data Based on Correlation-Predictive Deconvolution

Abstract

Ground-penetrating radar (GPR) is a widely utilized near-surface geophysical technique. However, the interpretation of GPR data remains challenging due to the presence of coherent noise, particularly multiples. This study investigates the autocorrelation profile characteristics of fundamental surface and internal multiples in velocity-increasing media. It then introduces a predictive deconvolution parameter selection strategy based on the energy distribution of primary waves and multiples within the autocorrelation profile, with the aim of simultaneously suppressing these multiples in zero-offset data. Subsequently, this strategy is applied to non-zero common offset data for both TE and TM polarizations. The results demonstrate that setting the prediction filter length equal to the number of single-trace sampling points, combined with a prediction step length ranging from the last PP events to the first PM events, effectively suppresses both surface and internal multiples. This approach significantly enhances the signal-to-noise ratio and improves the accuracy of profile interpretation in GPR common offset data, as evidenced by field data validation.