Abstract

Low Frequency Passive Seismic (LFPS) method is based on analysis of the micro-tremors generated from natural seismicity of earth. They are extremely low amplitude & low frequency signals. Spectral attributes like Ratio of Vertical and Horizontal amplitudes (V/H), Power Spectral Density (PSD), Peak Frequency Distribution (PFD), Frequency Shift of Maximum Spectral Peak, Polarisation etc. are generated during LFPS data processing and analysis. Anomalous response like relative increase in V/H ratio in 2-4 Hz range is observed over hydrocarbon pools. Besides micro-tremors, various events observed on the spectrogram obscure subtle signals linked to hydrocarbon reservoirs. They include Ambient, Anthropogenic and noises from local or distant sources with unsteady spectrum, short term/limited time perturbations, coherent and persistent noise streaks etc. These may be harmonic, quasi-harmonic, coherent or random in nature. Dominant horizontal and vertical streaks vary from area to area. These noise patterns need to be identified and suppressed /attenuated from the data before generation and analysis of spectral attributes. Deep learning approaches, specialized filtering, data analysis techniques, statistical methods help identify, characterize and suppress/attenuate noise patterns from the LFPS data. Noise suppression studies help to improve the quantitative estimates from anomalous attribute analysis, reduce the risk of misinterpreting the results thereby better risk mitigation and value addition.