Original Paper Investigation of Grade Bias Due to Core Loss Using Bivariate Conditional Distribution Babak Rajabinasab 1 and Omid Asghari 1,2 Received 9 August 2017; accepted 2 November 2017 Published online: 13 November 2017 Three-dimensional modeling of a mineral deposit was conducted based on the samples collected from the surficial and deep parts of the deposit. One of the factors that affect the quality of data is the recovery of cores acquired during exploration drilling. As the core recovery reduces, the grade in core or drilling mud increases. Core loss can introduce unpredictable errors and negative or positive bias into the mineral resources estimation. Investigating the grade–recovery relationship in a bivariate space helps to detect the grade bias. In an ideal state, there is no correlation between grade and recovery. In case the grade– recovery relationship possesses a negative or positive correlation, the grade bias is likely to be due to core loss. In practice, however, because there are many data in boreholes sample database, the grade–recovery relationship is not properly determined on a simple scatter plot of grade versus core recovery. Application of conditional distributions in a bivariate space provides the necessary tool for investigating the simultaneous variations of grade and core recovery. Therefore, having converted the grade and core recovery data into the normal space, the conditional grade distribution was determined for various recoveries, and then, the conditional grades expectations were determined and plotted against the recovery to investigate the grade–recovery relationship. Application of the proposed method in three different deposits showed that in low recoveries, due to waste loss, the grade values were overestimated, whereas in high recoveries, due to ore loss, they were underestimated. KEY WORDS: Mineral resources estimation, Core recovery, Grade bias, Conditional distribution. INTRODUCTION Mineral resources estimation, regardless of type and size of ore deposit, is vital to a mining project (Annels and Dominy 2003). To reduce the risk in mineral resources estimation, it is necessary to take samples from the surficial and underground parts of a deposit. Borehole drilling is essential but expen- sive to access directly to the ore minerals (Hassani- pak 2001; Committee 1997). Data obtained from exploration drilling are the basis of geological and grade modeling of a deposit (Annels and Dominy 2003). There are multiple error sources in drilling that can affect the accuracy of mineral resource estimation, including low core recovery (Annels and Dominy 2003; Hassanipak and Sharafoddin 2005; Moon et al. 2006; Haldar 2013). In practice due to geological and technical fac- tors, parts of a core are usually lost in core drilling (D2113-14 ASTM 2014). This problem extremely affects the aim of exploration drilling, which is direct access to representative samples. If the grade of the lost part is lower than the recovered one, positive bias will be introduced and the grade will be over- 1 Simulation and Data Processing Laboratory, School of Mining Engineering, College of Engineering, University of Tehran, Tehran, Iran. 2 To whom correspondence should be addressed; e-mail: o.asghari@ut.ac.ir 29 1520-7439/18/0100-0029/0 Ó 2017 International Association for Mathematical Geosciences Natural Resources Research, Vol. 27, No. 1, January 2018 (Ó 2017) DOI: 10.1007/s11053-017-9358-z