This study investigates whether pixel-based geostatistical modelling methods can be used to generate high net:gross ratio models with low connectivity of the net facies, a common combination in many natural geological systems. Connectivity as a function of net:gross is systematically measured in stationary, representative, horizontally isotropic three-dimensional models generated using three different pixel-based geostatistical methods: sequential indicator simulation, truncated Gaussian simulation and multiple-point statistics modelling. All methods are found to have percolation thresholds similar to, or substantially lower than, that of a random object-based model in which the net facies becomes macroscopically connected at a critical net:gross ratio of 0.27. A geometrical transformation previously defined for object-based models known as the compression method has been adapted to deal also with pixel-based models. Application of the method, calibrated using the newly established percolation thresholds, allows construction of geologically realistic facies models using a range of pixel-based methods. The resultant models contain quantifiable levels of connectivity that are defined independently of the model net:gross ratio.

这项研究调查了是否可以使用基于像素的地统计建模方法来生成高净：毛比率模型，而净相的连通性却很低，这是许多自然地质系统中的常见组合。在使用三种不同的基于像素的地统计方法生成的平稳，代表性，水平各向同性的三维模型中，系统地测量了作为net：gross函数的连通性：顺序指示器模拟，截断的高斯模拟和多点统计建模。发现所有方法的渗漏阈值都类似于或大大低于基于随机对象的模型的渗漏阈值，在该模型中，网状面在宏观的净毛比为0.27的情况下成为宏观连接。先前为称为压缩方法的基于对象的模型定义的几何变换已被修改为还可以处理基于像素的模型。使用新建立的渗流阈值进行校准的方法的应用，允许使用一系列基于像素的方法构造地质上现实的相模型。生成的模型包含可量化的连接级别，这些级别的定义独立于模型的净值：总值之比。