KANZI
EA
Prospectivity

li-v1.4

Continental brine lithium, Makgadikgadi

All runs Ranks ground
Mineral system
Li Lithium Definition 1.3
Continental brine lithium, Makgadikgadi
Signed by M. Mokgatlhe, MAusIMM CP on 22 Aug 2026
Physics-constrained ensemble, brine residence model · 500 m
Ran 02 Sept 2026, 170 min
Run by: Amini modelling team · Accepted by: M. Mokgatlhe, MAusIMM CP
Supersedes li-v1.3
Capture efficiency
0%25%50%75%100%0%25%50%75%100%randomBlind-site hit rate 0.75

The top 10% of ranked ground holds 66% of the known deposits.

Search-space reduction 94.3%

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

Validation
Blind-site hit rate 75%
Threshold agreed at MS1 50%
Block cross-validation 5 folds, 60 km blocks
AUC 0.81 ± 0.07
Physical constraints
  • Prediction is confined to closed sub-basins with a modelled evaporite sequence; drainage that leaves the basin cannot score high.
  • Graben-bounding structures set the lateral extent: brine is not predicted across a basement high.
  • Aquifer units shallower than the modelled palaeo-water table are excluded.
Features and importance
FeatureCritical processSystem roleImportance
Modelled evaporite thicknessTrap: evaporative concentration of brine in a closed basinTrap
33%
Gravity-derived basin depthTrap: accommodation space holding the brine columnTrap
24%
Distance to graben-bounding faultTransport: structural control on inflow and residenceTransport
18%
Felsic catchment shareSource: lithium-bearing volcanics weathering into the basinSource
15%
Playa persistence, multi-yearPreservation: a surface that has not been flushedPreservation
10%
Inputs
DatasetVersionConfidence filterCoverage
Makgadikgadi borehole hydrochemistry2025 compilationB or better, screened intervals44%
National gravity2021A and B only83%
North-west radiometrics1997C or better, levelled52%
Sentinel-2 evaporite and playa indices2026-08cloud < 5%98%

Makgadikgadi borehole hydrochemistry covers 44% of the modelled area and is the layer that governs the bounds.

Validation
Withheld siteWithheld byClassIn the high class
BS-LI-01 Sua Pan eastM. Mokgatlhe, MAusIMM CP High In the high class
BS-LI-02 Ntwetwe centreM. Mokgatlhe, MAusIMM CP High In the high class
BS-LI-03 Nata delta marginM. Mokgatlhe, MAusIMM CP Medium Missed
BS-LI-04 Mosu escarpmentM. Mokgatlhe, MAusIMM CP High In the high class
Outputs
  • Prospectivity surface 1.4 · COG, 500 m
  • Confidence surface 1.4 · COG, 500 m
  • Feature-importance maps 1.4 · COG per feature
  • Difference against li-v1.3 1.4 · COG, 500 m

Parameters

ensemble
3 members, physics-weighted
classCut
high ≥ 0.74, medium ≥ 0.50
residenceYears
12000
negatives
basin margin, outside closed drainage
Limitations and completeness
  • Borehole hydrochemistry covers 44% of the basin and is clustered along the Nata road.
  • Brine residence is modelled, not measured: the bounds are wide in the centre of Ntwetwe.
Against the previous edition
li-v1.3li-v1.4Change
Blind-site hit rate50%75%+25% Improved
Capture at 10% of area55%66%+11% Improved
Block-CV AUC0.760.81+0.05 Improved
AUC spread across folds0.090.07-0.02 Improved
High-class ground2,640 km²2,100 km²-540 km² Improved
Search-space reduction92.9%94.3%+1.5% Improved
Grid resolution5005000 Unchanged

Inputs that arrived

  • North-west radiometrics · 1997

Inputs revised

  • Makgadikgadi borehole hydrochemistry: 2024 compilation → 2025 compilation
  • Sentinel-2 evaporite and playa indices: 2026-05 → 2026-08

What moved the model

  • Felsic catchment share 0% → 15%
  • Playa persistence, multi-year 16% → 10%
  • Gravity-derived basin depth 27% → 24%
  • Distance to graben-bounding fault 21% → 18%
  • Modelled evaporite thickness 36% → 33%

Difference against li-v1.3 COG, 500 m

Targets from this run

2 targets are ranked from this run.

TargetScoreClassCompetent person
MKG-LI-003 Sua Pan North Brine0.78 High M. Mokgatlhe, MAusIMM CP
MKG-LI-007 Ntwetwe Pan Margin0.61 Medium M. Mokgatlhe, MAusIMM CP