Figure 1. Swath plots in Easting, Northing, and Elevation. Black: Real Data, Red: Kriging, Blue: GeaAI. Figure 1 illustrates the model's performance in the Northing direction. A significant high-grade shoot is observed around coordinate 7,004,500. Ordinary Kriging (red line) severely smooths this peak, underestimating the grade by over 50%. The Hybrid model (blue line) accurately tracks the peak, demonstrating its ability to resolve high-frequency spatial variation without overshooting in low-grade zones. Quantitative analysis of the Swath data shows that in the top 20% of high-grade zones, Kriging introduces a negative bias of -20.88 g/t, while GeaAI reduces this bias to -5.15 g/t, a 75.4% reduction in conditional bias. 4. DISCUSSION 4.1 Economic Implications: Recoverable Resources The smoothing effect of Kriging has direct economic consequences: it dilutes ore, lowering the estimated grade below the cut-off, effectively "destroying" economic ore in the block model. Figure 2. Grade-Tonnage Curves comparing metal recovery. Cut-off (g/t) Metal Error (Kriging) Metal Error (GeaAI) Value Recovery 124.1 -74.1% -24.2% +49.9% 134.4 -75.9% -24.7% +51.2% 144.8 -77.0% -22.2% +54.8% Table 3: Metal Recovery Analysis at Selective Mining Cut-offs At a selective cut-off of 144.8 g/t, the traditional Kriging model underestimates the contained metal by 77%. The GeaAI model recovers the majority of this value, with an error of only 22%. For a mining operation, this difference represents a massive increase in the Net Present Value (NPV) of the project, simply by using a better estimation algorithm that respects the selectivity of the deposit. 4.2 Operational Application: The "Value Window" We analyzed the model's performance across different drilling densities to determine the optimal application stage.
RkJQdWJsaXNoZXIy MTM0Mzk2