Kwekwe block 6

A ~116 ha four-corner block about 13 km south-south-west of the first Kwekwe block and 15 km south-west of Kwekwe town, forwarded by the holder on 2026-08-30 as four typed coordinate pairs. This is new ground for the study: no previously assessed site sits within 12.9 km of it. This is a where-to-walk assessment on free satellite data, not a claim that gold is present.

The zone check, this time confirmed by the holder's own screen. The coordinates arrived with no zone label, as at every site here. His GPS app displays the cursor in two zones at once, and reprojecting its 35S readout (777653 E / 7898026 N) into zone 36S returns a northing of 7896722 against the 7896721 shown on screen: a match inside one metre. The block is EPSG:32735, confirmed rather than inferred. Read as 36S it would land about 1,350 km east, in the Indian Ocean.
The block reads below its own neighbourhood; if any of it is worth pegging it is the eastern third. Whole-block median P(gold) is 0.174 at 10 m, the 19th percentile of the ground immediately around it — the second-weakest result of any block in this study, above only Kwekwe block 1 (8th). The terrain association is the one positive (64th percentile): the block sits on roughly the kind of ground local mines occupy, which is necessary and nowhere near sufficient. Everything the model does like is on the east side. The three best spots, P 0.575 at 778157 E / 7897914 N, P 0.542 at 778327 E / 7897484 N and P 0.501 at 778157 E / 7897644 N, all sit within about 300 m of the eastern corner, and the two eastern peg corners themselves score 0.38 against 0.14 and 0.18 for the two western ones. Walk the eastern edge first, starting at spot 1; the west half is background.
The footprint is approximate, and the verdict survives that. The four corners were supplied as an unordered list, and taken in the order given they trace a self-intersecting figure. They are polar-ordered here into a 116 ha convex quad. Scored under the alternative reading instead, the block reads a median of 0.206, the 31st percentile: a slightly better number, still below the local background median, and the same eastern spots carry it. The ordering ambiguity does not change what to do. The fix is one message: ask for a GPX or KML export from the app rather than typed pairs.

Every 10 m pixel over the block is inspectable: the heatmap is the fused P(gold) surface (dark = low, bright = high), and tapping anywhere shows the score, its rank against the local background and within the block, and every layer behind it, with the coordinate in peg format. The low ground is as visible as the high ground on purpose: control samples from dark pixels are what calibrate this model. Scores read surface geology, not grade.

The best-ground spots (UTM 35S). All six fall inside the block. Spots 1 to 3 (purple) are the eastern cluster and the first ground to walk; spots 4 to 6 (slate) trail off to the north-west at barely half the score:

The four supplied corners score the same way, and the gradient is the clearest thing on the block:

Corner UTM 35S P(gold)
a (west) 776844 E / 7897873 N 0.137
b (south) 777156 E / 7897191 N 0.179
c (north) 777771 E / 7898517 N 0.382
d (east) 778428 E / 7897503 N 0.377

For reference, the cursor position on the holder's own GPS screen, 777653 E / 7898026 N, scores 0.203.

The five checks behind that

Each is scored over the block against the ground immediately around it.

Check This block What it means
Terrain match 64th percentile the block sits modestly on the kind of terrain local mines occupy; terrain model calibrated on 52 mines within 25 km, LOO AUC 0.906, holding 0.836 under 8 km spatial cross-validation. Real skill, but coarse: it recognises greenstone-belt terrain type, not a deposit
Red-oxide (ferric) 57th percentile ferric ratio 1.528 vs background 1.511, marginally above the background median on ground that is 100% bare and 0% vegetated, so nothing is being masked. The number is not the point: mine separability sigma here is 0.092, and the 52 local mines sit at the 59th percentile themselves, so ferric-high does not mark gold in this district and this arm cannot rank the block either way
Structure / lineaments 45th percentile block-wide lineament density sits below the window mean (0.035 vs 0.063); dominant trend 120–135 deg, consistent with the belt grain. Roads, field edges and worked ground are the dominant false positives on bare terrain like this
ASTER alteration (SWIR) Al-OH 71st, carbonate/Mg-OH 71st, Mg-OH chlorite 25th percentile pre-2008 archive day scene (2004-07-30, 0% cloud), one of only five ASTER scenes here and the only usable low-cloud one. The Al-OH and carbonate highs look encouraging until the calibration is applied: the 51 local mines carry no alteration signal beyond chance (Al-OH sigma 0.102 at the 56th percentile, carbonate sigma −0.291 at the 43rd). At ASTER's 30 m the narrow orogenic alteration selvages here are sub-pixel, so the instrument has no discriminating power in this district and the block's elevation is uninformative
Regional P(gold) 0.174 whole-block median (19th percentile), p90 0.295, max 0.575 the combined estimate at native 10 m over 11,641 in-block cells; below the local background median, with the whole high end confined to the eastern edge

The neighbourhood, which is better than the block

Eighty-one known gold occurrences sit within 40 km. The nearest is 5.1 km away (204 kg produced at 16.3 g/t), and a 14.9 tonne producer at 2.9 g/t lies 6.8 km off, with an 8.6 tonne / 6.9 g/t mine at 11 km. That is a better-endowed neighbourhood than the first Kwekwe block had. It is also exactly the trap this study keeps documenting: district endowment is not block prospectivity. The regional model was built to separate those two statements, and on this block it separates them against the block.

Does the model recognise the producers next door?

Worth asking before trusting a negative verdict: two of the 81 known occurrences fall inside the 12.8 by 13.5 km window that was scored, so they can be checked directly against the same surface that judged the block.

At their catalogued coordinates they score 0.204 (the 204 kg / 16.3 g/t mine, 5.1 km away) and 0.233 (the 870 kg / 2.1 g/t mine, 5.6 km away): the 36th percentile of the window, near background. Taken at face value that would undercut everything on this page. It does not, and the reason is documented in the reference set itself: the Blenkinsop compilation is gridded to roughly half a kilometre, so a 10 m sample at the listed point is not a sample of the mine.

Widen to the ground around each point and the picture inverts:

Producer at the listed point best pixel within 250 m within 500 m within 750 m
204 kg at 16.3 g/t 0.204 0.445 0.474 0.747
870 kg at 2.1 g/t 0.233 0.748 0.860 0.880

Both producers have ground scoring 0.75 to 0.88, the top of the window, within 750 m of where they are catalogued, and the 870 kg mine's 750 m surroundings carry a median of 0.391 (89th percentile). The model does light up on real gold ground here; it is the coordinate that is coarse, not the surface.

That makes the block's reading harder to explain away rather than easier. Across all 116 hectares, no pixel exceeds 0.575 — the block has nothing resembling the 0.75-plus ground both neighbours sit on, and its median is at the 19th percentile while theirs sit near the top of the same window.

Compared to all blocks assessed

Same model run over each block:

Block Whole-block P(gold) Percentile vs local background
Belingwe-West 0.394 n/a
Silobela block 3 (assayed) 0.346 69th
Mvuma block 3 0.325 82nd
Zvishavane 0.282 67th
Kwekwe block 2, whole block 0.254 45th
Kwekwe partners block 0.229 43rd
Kwekwe block 6 (this), whole block 0.174 19th
Kwekwe block 1 0.129 8th

Only Kwekwe block 1 reads weaker. For scale, the Silobela ground whose walk-first spot assayed 2.33, 1.52 and 1.34 g/t on three of five grabs sits at 0.346 and the 69th percentile: twice this block's median. The gap between the block's median (0.174) and its best pixel (0.575) is the whole story here, and that best pixel is on the eastern boundary, not in the interior.

Data used

The P(gold) grid reads surface geology at 10 m. It tells you where to look on foot, not where a vein is. The useful next step is a walk along the eastern boundary: start at 778157 E / 7897914 N, work about 430 m south-south-east to 778327 E / 7897484 N, and take the 0.501 spot at 778157 E / 7897644 N on the way between them. Take grabs from the dark ground on the western half too. Low-scoring control samples are what this model is missing everywhere, and a handful from a block that reads at the 19th percentile is worth as much to the calibration as the targets are.