Definition
Proportion of model prediction errors attributable to geological domain boundary transitions versus within-domain prediction variance.
The Measurement Gap
When ore is sent to the waste dump or waste dilutes the ore stream, the financial impact is immediate and measurable. Traditional reconciliation metrics quantify the total ore loss and total dilution. What they cannot do is separate the sources: was the boundary wrong because the model drew it in the wrong place, or was it wrong because the geological contact is inherently variable at scales finer than the data density supports?
Legacy metrics like ore recovery ratio, dilution percentage, cost per ton, and even model bias factors treat all boundary errors as a single category. Physical Availability, MTBF, MTTR, and PM Compliance are equipment-focused metrics with no connection to geological boundary accuracy. Even traditional reconciliation factor analysis, which compares predicted vs. actual tonnage and grade by mining block, aggregates all error sources together. Without attribution, the geology team cannot determine whether the model needs improvement, the data density needs to increase, or the inherent geological variability simply cannot be resolved at the current drill spacing.
Decision Supported:
When ore is lost to waste or waste dilutes ore, how much is the model's fault vs. the geology's fault, and should the model's boundary predictions be trusted for dig-line decisions?
Formula
Plain Language: After a block is mined and assayed, compare the model's predicted ore-waste boundary to the actual boundary revealed by assays. Where the model was wrong, determine whether the error occurred in an area with sufficient data density (model error) or in an area where the data spacing was too wide to resolve the boundary (geological uncertainty). MBAR is the proportion of total misclassified volume that falls in the "model error" category.
Notation:
MBAR = Volume_model_error / (Volume_model_error + Volume_geological_uncertainty)
Ore_loss_rate_model = Tons_ore_to_waste_from_model_error / Tons_ore_total
Dilution_rate_model = Tons_waste_to_ore_from_model_error / Tons_ore_delivered
| Variable | Definition | Unit |
|---|---|---|
| Volume_model_error | Misclassified volume where the model deviated from assays in areas with sufficient data density to resolve the boundary | Cubic meters or tonnes |
| Volume_geological_uncertainty | Misclassified volume where assay spacing was too wide to resolve the boundary regardless of model quality | Cubic meters or tonnes |
| Tons_ore_to_waste_from_model_error | Ore tonnage sent to waste due to model boundary error (not geological ambiguity) | Tonnes |
| Tons_ore_total | Total ore tonnage in the reconciled block | Tonnes |
| Tons_waste_to_ore_from_model_error | Waste tonnage delivered as ore due to model boundary error | Tonnes |
| Tons_ore_delivered | Total tonnage delivered as ore | Tonnes |
Determination of model error vs. geological uncertainty: compare model error magnitude against kriging variance. Where model error exceeds what the estimation variance would predict, attribute to model error.
Proposed formula, pending validation pilots.