三维地质模型约束下地球化学场的Bayesian-MCMC推断

推断地球化学场元素的三维空间分布及其从深层向表层传输和分散的模式，对隐伏矿体三维定量预测具有十分重要的意义。当前化探数据的二维空间结构分析方法，难以揭示其空间分布特征及与地下地质结构之间的关系。项目拟从三维地质模型表达的地下空间场源、场作用及其介质出发，通过Bayesian-MCMC推断开展表生地球化学场的成因建模，并采用CPUs+GPUs异构并行技术使之在计算上可行。首先，根据不同岩相的地球物理响应，由物探数据推断从区域、矿床到矿体多尺度的三维物性模型，为地球化学场推断提供场作用介质；其次，根据矿床(体)的物质分异、蚀变分带，由岩矿采样数据推断原生地球化学场，为表生地球化学场推断提供场源；最后，将若干个场作用函数作为贝叶斯网络的候选节点，迭代找到与地表化探数据最优匹配的网络结构，推断原生与表生地球化学场之间的关联关系。项目的完成将为原生与表生地球化学场之间的场作用模式分析提供解决方案。

In the concealed orebody three-dimensional quantitative prediction, it is of great significance to infer the three-dimensional spatial distribution of elements in geochemical field and to probe elements transmission and dispersion patterns from subsurface to surface. At present, the two-dimensional spatial structural analysis method of geochemical exploration data is difficult to reveal the relationship between its spatial distribution characteristics and the subsurface geological structure. Starting from the field source, interaction and medium of subsurface expressed by three-dimensional geological model, this project intends to carry out reason modeling of surface geochemical field by Bayesian-MCMC inference, and to use CPUs + GPUs heterogeneous parallel technology to make it computationally feasible. Firstly, according to the geophysical responses of different lithofacies, the three-dimensional physical property models of multi-scale from region, deposit to orebody would be inferred from geophysical exploration data, which provide the field interaction medium for the geochemical field inference. Secondly, according to the law of material differentiation and alteration zoning of the deposit (orebody), the subsurface geochemical field would be inferred from the rock sample data, which provides the field source for the surface geochemical field inference. Finally, several field interaction functions would be taken as candidate nodes of Bayesian network, then the best network structure matching with the surface geochemical data would be found iteratively, that is, the correlationship between the surface and subsurface geochemical fields would be inferred. The completion of this project will provide a solution for the analysis of the field interaction patterns between the surface and subsurface geochemical fields.