A spectral method for reflectivity estimation

In this paper, a nonparametric method is used to estimate reflectivity from the seismic partial spectrum. Contrary to the conventional Wiener filter and the developing spectral inversion algorithm, this method can obtain unbiased and higher resolution results without the need for prior information related to formation reflectivity. The reflection amplitude with noise is estimated from each sampling time independently, the statistical properties of interference from noise and reflection at other sampling times are simulated simultaneously, and an adaptive filter for compressing these interferences is designed to get the estimation of the real reflection amplitude. This spectral method has advantages such as statistical stability and high resolving power. We first verify the feasibility of this method through numerical modelling experiments, and utilize Monte Carlo simulations to analyse its resolution, bias and variance properties statistically. We also apply this method to real seismic data, and find that the total resolution is improved significantly while the horizontal continuity is preserved; furthermore, some original geological features are highlighted. (paper)