考虑激电效应大地电磁贝叶斯反演-地质工程专业毕业论文.docx

万方数据 万方数据 Abstract The magnetotelluric method using natural electric field as its source, underground polarization body will produce the effect of induced polarization under the action of the electromagnetic field, and thus a significant impact on the magnetotelluric anomaly. Cole-Cole model can be used to describe the induced polarization of MT, the four parameters in the model(c,m,τ and ρ0 also knowns as spectral parameters) has its own physical significance. If we can extract the spectrum parameters from the magnetotelluric data,the IP anomalies can be used to distinguish the ore body when the resistivity variance of surrounding rock is not obvious. In this article, we presented the forward calculation of the magnetotelluric method for horizontally layered medium with the IP effect, and analyzed the influence of apparent resistivity on every spectral parameter of Cole-Cole model, and analysised the feasibility of extracting the IP information through the measured apparent resistivity data.. The IP information in the data of the magnetotelluric method is very week and exits serious nonlinear problem in inversion. General inversion of manetotelluric method is based on the deterministic theory, and it is difficult to obtain the real spectral parameters. This paper consider to inverting IP spectrum parameters through combining the method of Bayesian statistics with Stochastic inversion algorithm, and analysis the uncertainty of inversion . In the Bayesian inversion, multidimensional posterior probability density function (PPD) contains all the solution information of an inversion problems, we need to optimize the best initial model before extracted model information from PPD, and to sample and integral the model information in the whole high dimensional model space. Monte Carlo method that based on Markov chain can solve the problem of multiple integral calculation, constructing a smooth Markov chain with Gibbs distribution in a constant temperature to sample the mod