The Selective Mining Unit (SMU) is a nominal regular block volume that represents the scale of production that will be achieved in future mining. The vertical size of the SMU is almost always the bench height. The horizontal size varies based on geological, engineering, operational and other considerations (see below). A too-small SMU size would overestimate selectivity and underestimate dilution; conversely, a too-large SMU size would underestimate selectivity and overstate dilution. The SMU is a long-range resource modeling concept that helps determine the appropriate degree of grade smoothing in the resource model.
The SMU is not relevant for short-range or grade control modeling. At the time of grade control, it is recommended to create a high-resolution model of all relevant grades, translate them to the post-blast muckpile, optimize the destination of the mined material, and mine practical grade control volumes using best operating practices. It would be incorrect to model SMU-scale blocks for grade control and attempt to mine them; some boundaries could be delineated more precisely, and individual SMUs could not be mined independently.
Matheron described the theory of block variances in early publications on geostatistics (see, for example, page 1253 of "Principles of Geostatistics", Economic Geology, Vol 58, 1963, pp. 1246-1266). The concept of selective mining units is prominent in early references such as Mining Geostatistics (see, for example, page 488) and Geostatistical Ore Reserve Estimation (see, for example, page 306). Two foundational principles of geostatistics are (1) spatial correlation - quantified by the variogram, and (2) block volumes larger than the data support - quantified by average variogram values. Virtually all modern geostatistical references include discussion on change of support and selective mining units (including the paper by Ana Chiquini and I entitled "Mineral Resources Evaluation with Mining Selectivity and Information Effect", Mining, Metallurgy & Exploration, Vol 37, 2020, pp. 965-979).
The concept of an SMU helps calibrate the long-range resource model. The selectivity that would be achieved at the scale of composites is wildly optimistic. Smooth grade estimates for very large blocks based on widely spaced resource drilling would be distressingly pessimistic. For a chosen SMU scale, the selectivity could be theoretically predicted by change of support (Geostatistics Lessons - Change of Support and the Volume Variance Relation), then kriging could be calibrated to mimic the anticipated selectivity (Geostatistics Lessons - Introduction to Choosing a Kriging Plan).
There are many details to consider, but the most important parameter is the size of the SMU. The guiding principle is that the estimated tonnes, grade and metal from change of support with the chosen SMU size should match production at the time of mining. The actual production selectivity is sensitive to (1) geological considerations such as the nature of contacts, visual characteristics of the ore and fragmentation properties; (2) engineering decisions such as loading equipment, bucket width, mining direction, blast design; and (3) operational realities such as production data spacing, data quality and timing, estimation of the grade control model, dig limit mapping, supervision, and operational skill. The SMU size replaces all these considerations by a nominal size that reflects "selectivity".
Choosing the SMU size should be done carefully by (1) reconciliation with production data, if possible, (2) a detailed simulation study, if feasible, or (3) a decision by an experienced modeler, if necessary. The experienced modeler would consider their encounters with 10s of cases where different SMU sizes were chosen and, ultimately, reconciled with production. I support this third option; however, an expert's judgement is unavoidably subjective.
The use of a detailed simulation study is described in "Optimal selection of selective mining unit size for geostatistical modeling" a paper Oy Leuangthong, Steve Lyster, Chad Neufeld and I presented at the MININ conference in Santiago back in 2004. A high-resolution reference realization is constructed. The reference is sampled at the production spacing with appropriate error, a grade control model is built, dig limits are modeled, and the recovered tonnes, grade and metal are calculated. Then, the reference image is averaged to different SMU sizes to find the SMU size that matches the anticipated production. The practice has evolved to include more realistic assumptions regarding blast movement and variations in as-mined versus planned.
The preferred method to determine the SMU size is reconciliation with actual production data. The key assumption here is that the engineering, geological and operational considerations are similar to the volume being predicted. There are two main approaches (1) simulation, or (2) estimation. As you might expect, I like the simulation approach but would also consider estimation in practice. In both approaches, the starting point is to define reasonably large production volumes and calculate the actual tonnes, grade and metal produced in each volume.
The simulation approach would amount to simulating within the same large production volumes based on the available resource drilling. The model is then averaged to different SMU sizes to find the size where the predicted resources match actual production.
The estimation approach would consider resource drilling and estimation models. Different search strategies of increasing number of data would be considered until the estimates match actual production. Then, the SMU size that achieves the same smoothing of estimates would be inferred from change of support. The analytical change of support model is useful if the estimation is going to proceed to areas with different data spacing; otherwise, the tuned search strategy could be used directly.
The SMU size is used in long-range resource modeling. In estimation, the search plan is adjusted so that the smoothness of the model matches the analytical predictions of change of support. In simulation, the SMU dictates a higher resolution grid of simulated points that are subsequently averaged into SMU-sized blocks. There are times, however, when the complexity of the mining method combined with significant local variations in geological, engineering and operational considerations compromises the use of a constant SMU size. Nevertheless, in almost all circumstances, considering the SMU to describe future mining selectivity is practical and widely used.
The SMU size is not necessarily the block size for modeling (see next blog). The SMU size is a nominal volume that would anticipate future mining and provide the best possible resource estimates. Resource Modeling Solutions provides the software tools and workflows to determine the optimal SMU size and help resource estimators build the best possible model.

