LAUSR

Detection of fracture properties can be implemented using azimuth-dependent seismic inversion for optimal model parameters in time or frequency domain. Considering the respective potentials for sensitivities of inversion resolution and anti-noise performance in time and frequency domain, we propose a more robust azimuth-dependent seismic inversion method to achieve fracture detection by combining the Bayesian inference and joint time–frequency-domain inversion theory. Both Cauchy Sparse and low-frequency constraint regularizations are introduced to reduce multi-solvability of model space and improve inversion reliability of model parameters. Synthetic data examples demonstrate that the frequency bandwidth of inversion result is almost the same for time, frequency and joint time–frequency domain inversion in seismic dominant frequency band using the noise-free data, but the frequency bandwidth in joint time–frequency domain is larger than that in time and frequency domains using low- signal-to-noise-ratio (SNR) data. The results of cross-correlation coefficients validate that the joint time–frequency-domain inversion retains both the excellent characteristics of high resolution in frequency-domain inversion and the advantage of strong anti-noise ability in time-domain inversion. A field data example further demonstrates that our proposed inversion approach in joint time–frequency domain may provide a more stable technique for fracture detection in fractured reservoirs.