Inversion of multilayer amplitude versus offset data

A multilayer inversion method has been developed to estimate the P- and S-wavc velocities and densities from amplitude versus offset (AVO) data. It is assumed that the layers arc linear clastic, homogenous, isotropic, and horizontally bedded. The method uses the Zoeppritz equations as a physical model, generalized linear inversion (GLI) as a mathematical tool, and AVO data from common-mid-point (CMP) gathers as input. Output is the estimated seismic parameters. AVO data is corrected for spherical divergence, transmission, inelastic attenuation, vertical geophone response, and sourcearray receiver response losses before the inversion. The seismic parameters of the first layer are taken as a priori information, and parameters of the other layers are estimated downward recursively. Two different algorithms have been developed and tested using model AVO data with and without noise. Tests indicate that high signal to noise (S/N) and far offset to target depth (O'D) ratios are critical for satisfactory results. Other error sources in the estimated parameters arc inaccurate amplitude corrections, first layer parameters as a priori information, and sensitivity of the GL1 method to different parameters via the Jacobian matrix. Real data applications indicated that properly selected first layer parameters and, depending upon the algorithm used, a scaling factor and/or root-mean-square (RMS) velocities are the most critical factors.