Zimbabwe ten-peg block

A provisional first read of ten coordinates supplied on 3 September 2026. The working footprint is 812.6 ha, about 37 km from the existing Zvishavane site. This page answers one practical question: which ground deserves the first reconnaissance traverse? It is not evidence that gold is present.

Three inputs still need confirmation. Point B was supplied as 768309 E / 7796660 N and is provisionally read as 786309 E / 7796660 N, correcting a likely digit transposition. The unlabelled grid is provisionally WGS84 / UTM 35S. The supplied A–J order self-intersects, so the boundary is polar-ordered and approximate. Confirm B, the datum/zone, and the boundary traversal before using these results in the field.

Bottom line: two regional models agree that this block resembles above-background prospective terrain. The terrain-and-structure model ranks its median at the 94.1st percentile of nearby ground; the fused terrain, structure, and AlphaEarth model ranks it at the 91.3rd percentile. Lineament density is elevated too (85.9th percentile; north-west quadrant 93.2nd). This is a surface-resemblance result, not a discovery or grade prediction.

The field decision

Start with two broad search areas, not point targets:

East-central / north-east corridor: around 787100 E / 7795650 N, searching a 300–500 m neighbourhood. This contains the fused model's top spot and nearby regional, structure, and ASTER maxima.
South-central corridor: around 786450 E / 7794100 N, searching a 250–400 m neighbourhood. This is the tightest four-layer spatial convergence.

The numbered green areas on the map are those reconnaissance starts. Purple points are the fused model's ranked spots. Toggle the other evidence families to see whether a target persists or depends on one sensor.

Ranked fused-model spots

All six points fall inside the provisional polygon. The probability-like value is a relative model score, not the probability that a deposit exists.

What each layer says

Evidence	Claim result	Field meaning
Regional terrain + structure	94.1st percentile	Strong resemblance to terrain and structure occupied by the regional reference occurrences
Fused model	91.3rd percentile	AlphaEarth-dominated surface resemblance agrees with the regional read
Lineament density	85.9th percentile	Candidate linear features are elevated; roads, field edges, and drainage remain false positives
ASTER carbonate/Mg-OH	74.4th percentile	Moderate relative spectral-group response in a 2006 archive scene
ASTER Al-OH	66.4th percentile	Mild relative spectral-group response
ASTER Mg-OH/chlorite	14.4th percentile	No broad positive chlorite-group response
Sentinel-2 ferric ratio	10.8th percentile	The claim is not broadly iron-stained relative to nearby ground

The models are related evidence, not two independent assays. Both share terrain and structure inputs; 88.8% of the fused model's fitted importance comes from AlphaEarth embeddings. The ferric arm had only 41.6% usable claim coverage after vegetation masking and only one nearby mine for calibration. The local DEM-terrain arm could not run leave-one-out validation with one positive occurrence inside its fixed 25 km radius, so that result is deliberately absent rather than widened after the fact.

Read the surfaces

The strongest fused zone follows the same elongated central feature picked out by the lineament extraction. That spatial agreement is why it earns a traverse; it does not establish that the feature is a fault or that it is mineralized.

What to record on the ground

At both corridors, photograph and geotag bedrock, float, quartz veining, lateritic cover, cultivated boundaries, drainage, and any workings. Record strike/dip where possible and take control samples from low-ranked ground as well as the bright zones. Those observations—not another satellite index—are the next evidence needed to calibrate this block.

The complete provenance, reproduction commands, limitations, and all generated outputs are tracked under research/mineral-prospecting/sites/zimbabwe-10peg/.