Multiple-point simulation is a commonly used method in modeling complex curvilinear structures. The method is based on the application of training images that are open to manipulation. The present study introduces a new data-driven multiple-point simulation method that derives multiple point statistics directly from sparse data using copulas and applies them in simulation of complex mineral deposits. This method is based on simplification of N-dimensional copulas by its underlying two-dimensional copulas and taking advantage of conditional independence assumption to integrate information from different sources. The method was compared to Filtersim, a conventional multiple-point geostatistical method, through two synthetic data sets. Reproduction of cumulative distribution function, variogram, N-point connectivity, and visual patterns were considered in comparison. The copula-based multiple-point simulation (CMPS) method was implemented using trivial parts (almost 4%) of the synthetic data to extract required statistics while Filtersim was performed by giving the target image (100% data) as training image. Despite overwhelming data use in Filtersim, the CMPS showed compatible results to it. Application to synthetic data indicated that the method is a promising tool in the simulation of deposits with sparse data. The CMPS were applied in the simulation of two mineral deposits: (1) a porphyry copper deposit and (2) a magmatic iron deposit.

多点仿真是复杂曲线结构建模中常用的方法。该方法基于可操作的训练图像的应用。本研究引入了一种新的数据驱动的多点模拟方法，该方法使用联结函数直接从稀疏数据中导出多点统计数据，并将其应用于复杂矿床的模拟。该方法基于通过其底层二维联结函数对N维联结函数的简化，并利用条件独立性假设来整合来自不同来源的信息。通过两个合成数据集将该方法与 Filtersim（一种传统的多点地质统计方法）进行比较。比较时考虑了累积分布函数、变异函数、N 点连通性和视觉模式的再现。基于 copula 的多点模拟 (CMPS) 方法是使用合成数据的琐碎部分（几乎 4%）来提取所需的统计数据，而 Filtersim 则是通过将目标图像（100% 数据）作为训练图像来执行。尽管 Filtersim 中使用了大量数据，但 CMPS 却显示出与其兼容的结果。对合成数据的应用表明，该方法是模拟稀疏数据矿床的一种有前途的工具。CMPS 应用于模拟两种矿床：(1) 斑岩铜矿床和 (2) 岩浆铁矿床。