基于贝叶斯克里金的地下空间多源数据建模

Geomodeling with Integration of Multi source Data by Bayesian Kriging in Underground Space

摘要:

利用贝叶斯克里金方法集成钻孔数据与地震解译数据估计地质层面的高程.将线性贝叶斯理论运用于克里金估计，把用于空间估计的数据分为2类，即把钻孔数据视为硬数据，把数据不确性较大的地震解译数据视为软数据，构建了一个考虑贝叶斯方法的地统计模型，用区域性变量理论研究这2类数据的空间变化特征.通过对某煤矿煤层表面高程的估计以及与普通克里金方法的估计值与误差的比较证明了贝叶斯克里金法能较好地考虑地质剖面和地震解译数据的不确定性，在地质层面高程的估计方面具有一定的优越性，是一种可行的考虑数据全局不确定性的多源数据集成建模方法.

Abstract:

A subsurface model is usually built by integrating multi source geological data such as boreholes, geological maps and seismic interpretations. However, uncertainties inherited in these data are rarely quantified in the modeling process. In this study, Bayesian kriging method is introduced to integrate multi source geological data and estimate formation surface elevations. In this method, linear Bayes theory is applied to kriging estimation. Geological data is classified into hard and soft data. Hard data refers to coal seam data with enough confidence, such as boreholes. Soft data refers to coal seam data with uncertainty such as geological maps, cross sections and seismic interpretation information. Areal variable theory is employed to analyze spatial variation of both hard and soft data. This method is applied to the coal seam modeling of a coal mine in China. The estimates and errors of surface elevations are compared with those obtained from ordinary kriging method. Results show that Bayesian kriging method gives better results in terms of giving smaller errors of estimation. Therefore, Bayesian kriging is a useful method to incorporate multi source geological information and quantify uncertainties of geological data.