3D stochastic inversion of gravity data using cokriging and cosimulation.

摘要

The purpose of this thesis is to present an inversion method based on a geostatistical approach (cokriging and conditional simulation) for three dimensional inversion of gravity data including geological constraints.;Cokriging is a method of estimation that minimizes the error variance by applying cross-correlation between several variables. In this study the estimates are derived using gravity data as a secondary variable and the density as the primary variable. In the proposed method, the linearity between gravity and density allows us to obtain a covariance matrix of densities using observed data, i.e., we adjust the density covariance matrix by fitting experimental and theoretical gravity covariance matrices.;To obtain various reasonable solutions in order to see the variability that can be expected from the density covariance model adopted, a geostatistical simulation algorithm is applied. The simulation algorithm used in this thesis is based on the FFT moving average (FFT-MA) generator. Then the simulations are conditioned using cokriging results. The proposed method is applied to two different synthetic models: (1) the dipping dyke; (2) a stochastic distribution of densities. Then some geological information is added as constraints to the system of cokriging. The results show the ability of the method in fast integration of complex a priori information in the form of covariance functions. The proposed method helps us to modify the lack of resolution at depth and reduce the sensitivity to noise. Increasing the amount of information as constraints also helps to improve the estimation of the density distribution especially at deeper depths.;Finally, the southwest flank of the Matagami mining camp is considered as real data. The best height for upward continuation is studied for generating the residual map. Then our inversion method based on cokriging is applied to these residual anomalies in order to estimate the density distribution in this region. The co-simulation map is presented and the probability map is plotted in order to have a better interpretation. The results of inversion and simulation methods are in good agreement with the geology of the studied region.