Impact of In-Situ Density Spatial Model Methods on Resource Tonnages in Highly Intruded Coal Deposits.

Coal resource tonnage estimates are reliant on spatial modeling of critical data including in-situ density. To model coal in-situ density, numerous methods are available including deterministic, geostatistical, and machine learning-based methods. In most studies the focus is on comparing differing spatial model method accuracy, however, in this study, the focus is on determining if different in-situ density spatial model methods materially impact resource tonnage estimates. This study used data from an active open-cut coal mine located in the Bowen Basin, Queensland Australia. The coal mine has extensive areas that are impacted by lamprophyre and dolerite intrusion. This study compared tonnage estimates derived from in-situ density modeling from four spatial model methods including inverse distance weighting, ordinary kriging, geographic quantile regression forest, and random forest regression kriging. This study found that, at local scale and in areas highly impacted by intrusion, variation in tonnage estimates between in-situ density model methods was up to 4.40%. This variation was considered material as it was higher than the error of each spatial model. It is, therefore recommended that, especially in coal deposits which are impacted by intrusion, spatial model methods are carefully selected and evaluated. [ABSTRACT FROM AUTHOR]