De-noising and clean reconstruction of seismic signals using Space Lagged Singular Spectral Analysis (SLSSA)

Seismic records represent superposed oscillations of non-stationary and nonlinear signals shaped by the chaotically evolved crustal structures. Such seismic signals are generally assorted with correlated noise and this leads to serious problems in understanding their physical interpretation. We devised and employed a Space Lagged Singular Spectral Analysis (SLSSA) based algorithm to effectively de-noise and enhance seismic signal strength (amplitude) in order to clearly identify concealed layered structures. The SLSSA technique involves reducing rank order of the Hankel matrix by restricting low Eigen values that arise from noise and missing data. The basic noise reduction procedure involves (is performed by) alleviating contributions of fluctuating low Eigen values that arise from spatially uncorrelated noise processes. This is followed by reconstruction of the spatial data series applying appropriate weights based on Eigen values. This rank order reduction procedure does not require any prior information of the noise process. To test the efficacy of this new technique, we first applied the method to synthetic data contaminated with various kinds of noises of known percentages and missing data. Finally, guided by the above theoretical experiment, this method was applied to high resolution seismic reflection data of 60 channels, (8000 samples per channel) with 0.25ms sampling interval of 2s recording time. The new SLSSA algorithm successfully removes noise besides recovering missing signatures. Our analyses of actual seismic signals reveal some significant coal bearing horizons that were hitherto unmapped. Identification of these coal bearing horizons corroborates well with the geological prognosis for the study area. Key Words: Chaotic, Eigen values, Space Lagged Singular Spectral Analysis (SLSSA), Stochastic.