Prestack seismic inversion in time–frequency hybrid domain with adaptive prior constraint

Abstract Prestack seismic inversion serves as a critical method for predicting elastic parameters or physical property parameters of subsurface media. Conventional prestack inversion methods are mainly carried out in the time domain, usually with relatively low vertical resolution. Also, it is usually assumed that the prior information conforms to some specific probability distribution, which is difficult to express the change of reflection coefficient distribution caused by the lateral change of geological characteristics. To solve these problems, we present a prestack seismic inversion method in the time–frequency hybrid domain with adaptive prior constraint, which achieves elastic parameter prediction with improved resolution and precision. First, the advantageous frequency band is selected according to the signal-to-noise ratio of seismic spectra, which can enhance the vertical resolution of inversion results. Then, by introducing adaptive prior constraints, we incorporate the prior information on reflection coefficients of elastic parameters into the objective function. This circumvents assumptions regarding the global distribution of reflection coefficients while achieving enhanced characterization of lateral heterogeneity. We use 1D and 2D theoretical models and synthetic data to assess the new method’s validity and performance, and then apply it to a field data set to further examine its performance. Results indicate that it can achieve the estimation of compressional wave velocity, shear wave velocity, and density with high resolution and accuracy, while enabling good delineation of reservoir lateral continuity. Furthermore, the method exhibits strong noise immunity, ensuring reliable performance even in low SNR cases.