Modelling analysis of the limitation of P-wave AVAZ inversion

The limitation of P-wave amplitude versus azimuth (AVAZ) inversion for fracture attribute parameters was studied by numerical modeling. First, seismic reflection data from multiple azimuths and incident angles were synthetized using anisotropic elastic wave equation modeling performed on a geological model, with a homogeneous isotropic medium overlying an anisotropic medium comprising vertical fractures. Then AVAZ inversion for fracture attribute parameters was performed on the synthetic data, from noise free to a signal-to-noise ratio (SNR) level of 2. The results were analyzed by comparison with the actual attribute parameters of the model. We demonstrate that the fracture azimuth attributes from the AVAZ inversion of noise free data agree well with the actual models, with difference <1°, over incident angle range from 9.8° to 44.2°. Inversion results from noisy data, on the other hand, vary widely and failed to characterize the actual azimuth, though an average of many inversions may still be able to approximate azimuth. In contrast, fracture intensity attributes from AVAZ inversion generally carry large inherent errors. The AVAZ attributes from noise free data underestimate the intensity, using incident angle range of 15° to ~26°, with errors ranging from 8% to ~92%. Using a lager incident angle, within ~26° to 35°, overestimates the intensity, with errors up to ~461%. The inherent errors are so large and variable that the fracture intensity attributes from AVAZ inversion are not suitable for use as direct quantitative interpretation of fracture intensity. However, the relative differences of fracture intensity attributes from AVAZ inversion qualitatively agree with the actual intensity differences and may be exploited as a qualitative measure of fracture intensity variation applied to a very limited range of incident angles. For noisy data, it is difficult to characterize fracture azimuth, not only by a single inversion but also by a statistical mean of multiple inversion results and an effective de-noising processing strategy is a critical step for fracture azimuth inversion.