215 Figure 3: Decision tree illustrating the simulation workflow designed for this experiment. During the first two levels, the hierarchical structure is identical; thereafter, the tree bifurcates into two branches. Table 3 – Input parameters Parameter Value Dataset Jura (Goovaerts, 1997) Variable Rock Categories 1, 2, 3, 4, 5 Global proportions 0.18; 0.40; 0.25; 0.02; 0.15 Grid cells (nx, ny, nz) 250; 295; 1 Cell size 17.5 x 17.5 x 1 Total nodes 73,750 Random seeds 69069; 69071; 69073; 69075 Max. data / nodes 12/12 Max. per octant 6 Search radii (hx, hy, hz) 1200; 1200; 1 Variogram model (same for all classes) Isotropic spherical Nugget effect 0.0 Sill 0.25 Ranges (hmax, hmin, v) 1000; 1000; 0 Kriging type Ordinary Indicator Kriging Hierarchy 1 2; 3; 5; 1; 4 Hierarchy 2 2; 3; 1; 5; 4 4. RESULTS The results are presented as a series of figures (Figures 4-8). Each figure shows two maps, one for each defined hierarchy, sample proportions ranked (left) and declustered proportions ranked (right), allowing a direct visual comparison of how alternative decision-tree orderings progressively allocate space. The rock categories are progressively allocated in space according to the defined hierarchies, following the decision-tree structure shown in Figure 3, with the corresponding step illustrated at the top of each map.
RkJQdWJsaXNoZXIy MTM0Mzk2