Thin-layer scaling effects on AVO modelling

Abstract

In AVO modelling with log data, it is a general practice to upscale the log data for reasons such as computational efficiency. In this study, we investigate the AVO sensitivity to thin-layer effects by upscaling log data at different scales. A series of Kennett-based reflectivity modelling exercises are conducted using the resulting elastic 1D models. In the presence of thin layers, it is essential to incorporate interbed multiples (including converted wave modes) and therefore, primary-only Zoeppritz modelling could lead to misleading results. On one hand, very fine scale models involve a complex wavefield tuning that renders the theoretical Zoeppritz plane-wave reflection coefficients meaningless. On the other hand, wavefield complexities due to thin-layer multiple scattering also introduce uncertainties in the modelled AVO behaviour. In contrast, with increasing layer scales, Zoeppritz plane-wave reflection coefficients become more applicable, and interbed multiple and tuning effects also have less influence upon the AVO behaviour. At such scales, primary-only AVO modelling by ray tracing where reflection amplitudes are determined by Zoeppritz equations could be an efficient alternative to full-wavefield reflectivity modelling. In our synthetic examples, this occurs at layer scales (d) satisfying a relation to dominant frequency (?) of ?/d = 4 ~ 5. However, ? and d are interdependent, since the effective velocity used to calculate wavelength ? depends on scale d. To resolve this dilemma in practice, surface seismic data could be analysed and a Q sensitivity analysis may be necessary to determine the dominant wavelength at different target levels.