Abstract:Covariance velocity spectrum is based on the fact that the sampled data can be arranged in eigen structure of covariance matrix. The data covariance matrix is obtained by performing NMO correction, moveout scan and residual moveout correction of CDP gathers. The eigenvalue of covariance matrix can be used to separate signal from noise. Signal covariance estimate may be derived from the dominant eigenvalue, and noise covariance estimate from subordinate eigenvalue, so that covariance velocity spectrum has higher resolution than conventional stack correlation and semblance coefficient velocity spectra. Such high resolution can be seen in t0 time direction, particularly in stack velocity direction. Because this method is used to improve the resolution of velocity spectrum, we must pay more attention to the following points. ·Time window is not longer than main energy duration of the wavelet which is analysed. ·Moving increment of time window is not longer than half time window, and ·Big velocity increment is avoided here.