A case study on top cut model for resource estimation of a fracture related gold deposit

Accurate gold reserve estimation and 3D resource modelling using blasthole data remains a significant challenge due to the presence of highly skewed, long-tailed grade distributions dominated by extreme high-grade values. This issue frequently arises in gold deposit grade estimation, particularly in fault-controlled gold deposits with high grade variability. With applying kriging estimation, there are challenges due to a tiny minority of extreme values with a significant proportion of the metal content and part of the data variability, potentially introducing estimation bias. Traditional approaches such as capping, have been employed to address this challenge, which can yield acceptable results in some mining operations, but exists underestimation risk. To improve the practices of handling long-tailed grade data, top-cut model (2013) was proposed. This method decomposes the grade data into three components: a truncated grade representing the mid-to-low grade population, a weighted indicator term representing the mid-to-high grade population, and a residual term capturing the extremely high-grade population. The estimation is based on the cokriging of the truncated grade and the indicator and a separate kriging for residuals. Using a fault-controlled gold deposit, Xincheng gold deposit in China as a case study, this paper evaluates the estimation performance of the top-cut model compared to ordinary kriging and capping through cross-validation and block model validation via trench data. The result shows that top-cut model provides the best estimates for the gold orebody with high spatial variety compared to other methods. Furthermore, it is revealed that the estimation of top-cut model effectively limits the spatial influence of outliers while preserving more variability of the grade distribution—crucial for realistic reserve estimation and informed mine planning.