A Quantitative Evaluation of Input Data Variables for Model-based PP Inversion Using Conjugate Gradient Algorithm

Using any model-based pre-stack inversion process requires three types of input data: wavelet, initial model and seismic gathers. In this paper we analyze the impact of the errors possible, such as wavelet estimation errors, initial model errors and seismic signal to noise ratio (SNR), using the non-linear conjugate gradient method for inversion to P-wave velocity (Vp), S-wave velocity (Vs) and density () data. Using this synthetic model response as the recorded seismic data, we compare the inversion results to the Vp, Vs and density from actual well data for three zones of interest. Improper rock properties from the inversion will lead to incorrect conclusions about the viability of a prospect. We arrive at conclusions about the impact of wavelet, initial model and seismic SNR required for acceptable inversion. We also offer suggestions for future research and development to increase the accuracy of the inversion results.