High-frequency recovery of surface seismic data from autoregressive spectral extrapolation and Q-compensation

Abstract

Summary Autoregressive spectral extrapolation and Q-compensation allow for the recovery of the high-frequency component of the wavelet spectrum that has been attenuated by the Earth filtering effect. Autoregressive spectral extrapolation utilizes a predictive function while Q-processing uses a backward propagation process to recovery high-frequency. Despite the fact that the two methods are distinctive they share a common goal that turns out be complementary. Here both methods are combined into a single algorithm. First, the higher end (~30-75 Hz) of the frequency spectrum that generally has a lower signal-to-noise ratio is predicted from the lower part (~10-35 Hz) of the spectrum having a higher signal-to-noise ratio using an autoregressive spectral extrapolation. Then, seismic traces are Q-compensated using attenuation values that are estimated from seismic data. The combination of both methods improves the temporal resolution. The increase of seismic energy in the high-frequency band can be up to 17% and on average 10% when compared to data that are solely Q-compensated allowing the resolution of more subtle seismic stratigraphic features.