Stochastic modeling of iron in coal seams using two-point and multiple-point geostatistics: A case study

Abstract

This work addresses the problem of quantifying iron content in a coal deposit in the Republic of Kazakhstan. The process of resource estimation in the mining industry usually involves building geological domains and then estimating the grade of interest within them. In coal deposits, the seam layers usually define the estimation domains. However, the main issue with the coal deposit in this study is that the iron dataset is solely based on data from three newly drilled drill holes located a significant distance apart and additional rock samples from stopes. A massive amount of geological information comes from legacy drill hole data sampled a long time ago, but there is no evidence of proper QA/QC being performed on those samples. For this reason, a workflow was introduced to construct a representative training image from legacy data and stochastically model geological domains within these three drill holes using a multiple-point geostatistics technique. Once the geological model was obtained, a two-point geostatistics algorithm was applied to model the iron inside each geological domain. The results showed that direct sampling (DeeSse) is a suitable multiple-point geostatistics algorithm that can reproduce the longrange connectivity and curvilinear features of seam layers. Furthermore, a sequential Gaussian simulation was used to model the iron in the corresponding domains. Both methods were extensively evaluated using different statistical tools and analyses.