Rock-Physics-Guided Time-Frequency AVO Inversion for Quantitative Estimation of P-wave Velocity Dispersion Attribute in Hydrocarbon Identification

Abstract

Extracting dispersion information related to the frequency-dependent behavior of wave propagation from seismic data has become an effective method for hydrocarbon identification. However, existing approaches primarily rely on conventional frequency-dependent amplitude-versus-offset (FD-AVO) inversion, which provides only qualitative evaluations of elastic dispersion gradients without fully incorporating rock physics insights demonstrated by laboratory measurements and theoretical studies. A rock-physics-guided time-frequency AVO (TF-AVO) inversion framework is presented to bridge this gap by integrating rock physics knowledge with seismic data to enable quantitative estimation of P-wave velocity dispersion. Log-derived frequency-dependent velocities obtained from rock physics modeling are incorporated into the seismic inversion as physical constraints, ensuring consistency between predicted dispersion attributes and rock physics responses. A broadband rock physics modeling method is first established, from which frequency-dependent P-wave velocities and corresponding dispersion attributes are simulated using logging curves under the constraints of model-based inversion results. The modeled dispersion attribute is calibrated against well logs to quantify their sensitivity to gas-bearing intervals. Then, the log-derived frequency-dependent velocities are integrated with spectrally decomposed post-stack seismic data using waveform indication simulation to construct initial time-frequency models for TF-AVO inversion. The inversion is performed in the time-frequency domain through the limited-memory Broyden-Fletcher-Goldfarb-Shanno (L-BFGS) optimization algorithm to quantify frequency-dependent P-wave velocities. Field applications demonstrate that the estimated P-wave velocity dispersion exhibits improved consistency with gas saturation logs compared to the FD-AVO results, supporting both the validity of the adopted rock physics modeling and the effectiveness of the TF-AVO inversion. The proposed framework establishes a robust approach based on physical constraints for quantitative dispersion estimation, enhancing the reliability of hydrocarbon identification.