Improving Vertical Resolution of Nonstationary Seismic Data by Spectral Modelling in Logarithm Time–Frequency Domain

Summary Conventional resolution improvement methods assume that the seismic wavelet is time-invariant, which means the seismic data is stationary. However, seismic wave attenuation and scattering make the seismic wavelet vary in the process of propagation. In this study, we provide a spectral modelling method to estimate the time-varying wavelet using Fourier series fitting in logarithm time-frequency domain. Firstly, the generalized S-transform is used to decompose each seismic trace, which provides a good time-frequency distribution for estimating the time-varying wavelet, and then convert it to logarithm time-frequency domain. Secondly, a higher-order Fourier series is used to fit the timevarying wavelet spectra at each time sample of logarithm time-frequency domain. Finally, we use the time-varying wavelet spectral to spectrally balance seismic data to flatten the seismic response and improve vertical resolution. We investigate the feasibility of the proposed method via a synthetic and field data example. The results show the good performance in improving the vertical resolution of seismic data.