KANZI
EA
Prospectivity

au-v1.1

Orogenic gold, Tati and Vumba greenstone belts

All runs Ranks ground
Mineral system
Au Gold Definition 1.0
Orogenic gold, Tati and Vumba greenstone belts
Signed by K. Dlamini, FGSSA on 05 Sept 2026
Gradient-boosted trees, greenstone shear model · 250 m
Ran 10 Sept 2026, 185 min
Run by: Amini modelling team · Accepted by: K. Dlamini, FGSSA
Supersedes au-v1.0
Capture efficiency
0%25%50%75%100%0%25%50%75%100%randomBlind-site hit rate 0.71

The top 10% of ranked ground holds 62% of the known deposits.

Search-space reduction 94.4%

The share of the modelled area a competent person can set aside for this commodity.

Validation
Blind-site hit rate 71%
Threshold agreed at MS1 70%
Block cross-validation 5 folds, 25 km blocks
AUC 0.82 ± 0.06
Physical constraints
  • Prediction is confined to greenstone lithologies and their 5 km structural envelope.
  • Granitoid interiors cannot score high without a mapped shear.
  • Alluvial workings are excluded from the label set.
Features and importance
FeatureCritical processSystem roleImportance
Distance to second-order shearTransport: shear zones focus auriferous fluidTransport
35%
Banded iron formation contactTrap: chemical trap at a reactive iron-rich hostTrap
30%
Magnetic texture contrastTransport: alteration destroys magnetite along the fluid pathTransport
20%
Hard-rock occurrence density, 10 kmSource: historical workings mark a fertile corridorSource
15%
Inputs
DatasetVersionConfidence filterCoverage
Geology 1:250k2023all, lineage recorded100%
Regional aeromagnetics, eastern tiles2004 levelled, 2026 infillB or better88%
Mineral occurrence register, hard-rock only2026-09B or better, re-labelled from 1:50k55%

Mineral occurrence register, hard-rock only covers 55% of the modelled area and is the layer that governs the bounds.

Validation
Withheld siteWithheld byClassIn the high class
BS-AU-01 Map NoraK. Dlamini, FGSSA High In the high class
BS-AU-02 Signal HillK. Dlamini, FGSSA High In the high class
BS-AU-03 Golden EagleK. Dlamini, FGSSA High In the high class
BS-AU-04 Monarch southK. Dlamini, FGSSA Medium Missed
BS-AU-05 Phoenix flanksK. Dlamini, FGSSA High In the high class
BS-AU-06 TekwaneK. Dlamini, FGSSA High In the high class
BS-AU-07 Bushman shearK. Dlamini, FGSSA Medium Missed
Outputs
  • Prospectivity surface 1.1 · COG, 250 m
  • Confidence surface 1.1 · COG, 250 m
  • Feature-importance maps 1.1 · COG per feature
  • Against historical exploration 1.1 · COG, 250 m
  • Difference against au-v1.0 1.1 · COG, 250 m

Parameters

learningRate
0.06
maxDepth
5
trees
750
classCut
high ≥ 0.76, medium ≥ 0.50
negatives
2:1, granitoid interiors and barren greenstone
Limitations and completeness
  • Magnetic infill closed the eastern gap to 88%; the north-east corner is still inferred.
  • Re-labelling removed 34 alluvial workings from the label set, which narrows the training population.
Against the previous edition

Trust changed between these editions: Does not rank to Ranks ground.

au-v1.0au-v1.1Change
Blind-site hit rate28.6%71.4%+42.9% Improved
Capture at 10% of area29%62%+33% Improved
Block-CV AUC0.640.82+0.18 Improved
AUC spread across folds0.130.06-0.07 Improved
High-class ground1,950 km²1,180 km²-770 km² Improved
Search-space reduction90.7%94.4%+3.7% Improved
Grid resolution2502500 Unchanged

Inputs that arrived

  • None

Inputs revised

  • Regional aeromagnetics, eastern tiles: 2004 levelled → 2004 levelled, 2026 infill
  • Mineral occurrence register, hard-rock only: 2026-07 → 2026-09

What moved the model

  • Occurrence density, 10 km 21% → 0%
  • Hard-rock occurrence density, 10 km 0% → 15%
  • Magnetic texture contrast 15% → 20%
  • Banded iron formation contact 26% → 30%
  • Distance to second-order shear 38% → 35%

Difference against au-v1.0 COG, 250 m

Targets from this run

1 targets are ranked from this run.

TargetScoreClassCompetent person
TAT-AU-011 Vumba Greenstone East0.73 Medium K. Dlamini, FGSSA