Early-stage gold exploration has an uncomfortable arithmetic problem. To find out whether a licence
block is worth drilling, the classical answer is to walk it, map it, cut trenches, run soil and rock-chip
geochemistry, and then drill — a sequence that consumes months of field season and a large fraction of a
junior’s treasury before anybody knows whether the ground deserved the attention. Gold anomaly
detection in Zambia from orbit inverts that order. It asks the satellite record first, narrows the
ground that merits a rig, and leaves the expensive, invasive work for targets that have already survived a
multi-sensor screening.

This case study walks through exactly that exercise on a real Zambian gold licence. We screened a
283.2-hectare block (699.8 acres), fused twenty-one orbital and geophysical data layers into a single
composite prospectivity surface at 10-metre ground resolution, extracted ten discrete economic anomaly
zones covering 64.2 hectares, and modelled a satellite-derived exploration target of roughly
11,450 to 14,900 troy ounces of contained gold at a tonnage-weighted head grade of
0.51 g/t Au. Then we did the part that matters more than the headline: we published which evidence layers
agreed with that signal, which ones disagreed, and what it would cost — to the metre and to the rig-day —
to find out who was right.

Want this run on your own ground? Get a
quote for a satellite mineral scan, or draw your area of interest directly on the map at
mining.farmonaut.com — our mineral-detection mapping
workspace.

“One zone in ten holds 30% of the gold — 4,482 ounces of the 14,896 ounces detected.”

A 283-hectare Zambian licence block, and the question it posed

Zambia is far better known for copper than for gold. That is precisely what makes its gold ground
interesting: large tracts of prospective terrain sit inside licences held for other reasons, or held
speculatively, with no modern exploration dataset behind them. The block in this study is one of those.
It is a working licence within a larger Zambian concession, held by a private mining company that came to
us with a straightforward commercial question — is there anything here worth a drill program, and if
so, where exactly do we put the first hole?

The physical setting is unusually cooperative. Mean elevation across the block is approximately
1,333 metres above sea level, and mean slope is 4.66° — comfortably below the 25° threshold at which
open-pit geotechnics start to get complicated, and well within the range where a rig can be moved and
levelled without earthworks. Terrain accessibility was rated High across the whole block. Minimum distance
to the nearest building is 74 metres, and drainage, while well organised, does not cross a single anomaly
polygon. In exploration terms this is easy ground: nothing about the site itself is going to be the reason
a program fails.

What the site does have is cover. The dominant surface context across the licence is dense
vegetation — miombo-style woodland over a laterite-influenced regolith. That matters enormously, because
cover is the single biggest enemy of optical mineral detection. A bare, arid outcrop broadcasts its
mineralogy to a satellite sensor. A vegetated, weathered surface muffles it. Any honest account of
gold anomaly detection in Zambia has to start by conceding that the country’s vegetated
interior is a harder detection environment than, say, the Arabian Shield or the Atacama — and then explain
what you do about it.

The commercial problem in one paragraph

A junior or private licence holder with a 283-hectare block faces a capital-allocation decision, not a
geological one. Walking and geochemically sampling that area to a standard that would justify drilling is a
multi-month, multi-crew exercise; at typical African field-cost rates it consumes a meaningful share of an
early-stage budget before a single metre is drilled. And the classical sequence gives you no way to
rank what you find until quite late. Satellite screening is attractive not because it replaces
drilling — it cannot and does not — but because it reorders the spend: cheap, fast, non-invasive
prioritisation first, expensive confirmation second, and only on ground that earned it.

How gold anomaly detection in Zambia actually works

Every mineral and every alteration assemblage interacts with electromagnetic energy in its own way.
Iron oxides absorb and reflect differently from clay minerals; silicified rock stores and releases heat
differently from vegetated soil; a fault-controlled quartz corridor scatters radar differently from
undisturbed regolith. A satellite gold exploration screen is, at bottom, the systematic exploitation of
those differences — reading many independent physical properties of the same patch of ground, then looking
for the places where several of them line up in a pattern consistent with gold mineralisation.

The sensor stack

This screen drew on twenty-one orbital and geophysical layers. The table below lists the primary data
sources and what each one contributes geologically. Figures in parentheses are each sensor’s horizontal
ground sample distance — the on-ground footprint of one pixel — not a depth.

Each layer is normalised and combined into a single per-pixel composite prospectivity score — a
weighted composite of multiple normalised sub-scores spanning alteration mineralogy, structural density,
surface evidence, thermal signature and geomorphological host criteria. Contiguous high-scoring pixels are
then grouped into discrete anomaly polygons by thresholded connected-component analysis, each fitted with
an oriented minimum-area bounding box so that strike, width and elongation can be measured. Every polygon
carries a coherence score that combines its elongation with its boundary smoothness, because real
vein-controlled targets are elongate and smooth-edged, while noise is blobby and ragged.

From anomaly polygon to tonnes and ounces

Turning a mapped polygon into a tonnage and a contained-metal figure is deliberately conservative
arithmetic, and it uses industry-standard parameters throughout so that any competent person can reproduce
or challenge it:

The prism factor of 0.7 is the deliberate haircut: it assumes a mapped polygon does not fill its own
bounding volume, and it is the single most important reason these numbers land where they do rather than an
order of magnitude higher. The 18% dilution allowance adds waste rock at the mapped grade to reflect what a
mining operation would actually move. Recovery is a blended 81% — conservative for an oxidised, near-surface
gold system, where heap-leach response typically runs 75–90% and conventional carbon-in-leach 88–94%.

Read more about the method behind these screens on our
satellite-based mineral detection
page — it explains how orbital spectral data becomes a ranked target list for your licence.

What gold anomaly detection found: 10 zones, ~11,450–14,900 oz

Ten discrete economic anomaly zones were extracted from the composite surface, together covering
64.24 hectares (158.74 acres) — 22.7% of the 283.2-hectare surveyed block. Every zone sits
above the 0.25 g/t Au economic cut-off. All ten were classified as gossan-zone mineralisation: an
iron-oxide-rich, oxidised surface cap over a sulphide-bearing system, which is exactly the style you would
hope to detect from orbit, because the oxidation cap is what the sensors can actually see.

Gold anomaly detection output for the Zambian licence block — bubble area scaled sub-linearly to contained gold, with grade and contained-metal labels per zone. Coordinate axes and the client’s name have been removed from the original figure; the north arrow and scale bar are retained. Contained-gold labels are the report’s central diluted estimates.

The distribution of that metal is heavily skewed, and that skew is the single most commercially useful
finding in the dataset. Two zones out of ten carry 54.6% of the contained gold. The largest
single zone alone accounts for 30.1% of gross in-situ value — 139.4 kg, or 4,482 troy ounces, at a modelled
0.500 g/t across an 858-metre strike length. For a licence holder deciding where to spend a first drill
budget, that concentration is worth more than any headline total: it means a two-target program can test the
majority of the block’s modelled value.

The master comparison table — all ten zones

This is the full inventory. Grade is the modelled diluted grade that the contained-metal totals are
built on; the adjacent column shows the pixel-level P10–P90 grade spread, which is what sets the honest
uncertainty band on the portfolio. Tonnage, strike, strip ratio and value all reconcile to the portfolio
totals in the final row.

Total diluted ore tonnage across the ten zones is 909,224 tonnes (1,002,246 US short tons). Per-zone
strike lengths run from 89 metres up to 858 metres, and strip ratios sit in a tight band from 2.60 to 2.86,
weighted average 2.74 — comfortably inside the 3.0–4.0 range that surface-amenable gold operations routinely
accept. The uniformity of those strip ratios is itself informative: it says the ten zones share a single
depth architecture rather than representing ten unrelated geological situations.

Per-pixel modelled gold grade across the anomaly footprint. Pale cream is roughly 0.3 g/t Au, mid-orange around 0.6 g/t, and the isolated deep-red cells exceed 1.0 g/t. Coordinate axes, the licence outline and the colour bar have been removed for confidentiality — the polygon-averaged grades in the master table collapse this heterogeneity to one number per zone, which is exactly why the P10–P90 spread matters.

That per-pixel view is worth dwelling on, because it is where the honest uncertainty lives. A single
grade figure per zone is a convenience for a table; the underlying model produces a distribution. Peak
modelled grade anywhere in the block reaches 0.609 g/t at polygon level, with individual pixels running
above 1.0 g/t, while the P10 of several zones sits below 0.40 g/t. Publishing a single point value for a
satellite-derived estimate implies a precision the method does not have — which is why every resource and
economic figure in this case study is quoted as a band.

Where the value sits

Reading the evidence: which signals agreed, and which didn’t

This is the section most satellite exploration marketing leaves out, and it is the section that
determines whether a report is useful. A composite prospectivity score is a rollup of independent evidence
layers. If you only publish the rollup, you hide the disagreement. So the delivered report opens the
composite back up: each anomaly carries eight independent signal families, and each family is scored Strong,
Moderate, Weak or Marginal — both per zone and across the whole block. Credible gold anomaly detection lives
or dies on this table, not on the headline ounce count.

The pattern is unambiguous, and it is not entirely flattering. Thermal anomaly fired Strong on
all ten zones. Vegetation biogeochemistry fired Strong on five and Moderate on three. Radar surface
roughness fired Strong on four and Moderate on six. Those three families are the load-bearing evidence here,
and they are precisely the families that work under vegetation cover.

But hydrothermal alteration returned Marginal on all ten zones, and magnetic signature returned
Marginal on all ten. Block-wide mean ASTER alteration composite is effectively zero. That is two of
eight independent evidence layers declining to corroborate the signal. The report does not bury that; it
publishes it as a formal risk flag against every polygon, with a named mitigation.

“Thermal sensors fired Strong on all ten zones. Magnetics fired Marginal on all ten.”

The three detection-risk flags, and what resolves each one

Not one of the ten polygons raised zero flags. Combine that with the composite signal class — every zone
came back Moderate, none Strong — and the classification rule produces an unambiguous
verdict: zero high-confidence zones, zero medium-confidence zones, ten low-confidence zones.
All 14,896 modelled ounces and all USD 60.4M of headline in-situ value sit in the low-confidence tier.

Per-pixel signal strength class across the anomaly footprint. Amber cells are Moderate class; the two tiny green cells are the only Strong-class pixels anywhere in the block. Axes, licence outline and legend removed for confidentiality. This figure is the reason the confidence table reads as it does — there is simply no Strong-class population to promote.

It is worth being precise about what “low confidence” does and does not mean here. It does not mean the
anomalies are false. It means the surface evidence is partial, and that ground-truthing must precede any
drill commitment. The tier label governs the order in which zones are validated and how much capital
is committed at each phase — it does not gate whether validation happens. All ten zones enter the Phase-1
validation sequence. A further safeguard also ran and came back clean: a pixel-consensus check that demotes
any zone whose polygon-average strength class overstates its in-pixel evidence found no drift on any polygon,
so the class labels in the table reconcile with the underlying pixel distributions.

The screening funnel — from licence to lead targets

From anomaly to drill collar: depth, geometry and a costed program

A prioritised target list is only actionable if it comes with geometry. Every zone in this dataset
carries a modelled depth profile, a recommended drill azimuth and inclination, a hole length and a hole
count — the parameters a drilling contractor needs in order to quote. This is where gold anomaly detection
stops being a map and starts being a budget line.

The depth architecture across the block is remarkably consistent. Top of ore sits at
91 metres below surface across all zones, with the deepest modelled base at
135 metres (298 to 442 feet). Both figures are vertical distances measured downward from
local ground surface, and both sit well inside the practical depth limit for open-pit extraction of an
oxidised gold system, which typically runs to around 200 metres. Combined with the tight 2.74 weighted
strip ratio, that makes every one of these zones nominally open-pit amenable.

3D drill target vectors rendered over the terrain model. Green markers are proposed collars at each anomaly centroid, blue lines the planned hole paths at their recommended azimuth and −70° inclination, and orange diamonds the projected piercement point at target depth. Axes are relative to the block centre and the terrain extent has been cropped; the delivered client version is fully georeferenced.

The Phase-1 drill program, costed

The recommended full program is 76 HQ diamond core holes distributed across the ten zones, weighted
towards the two lead targets — 15 holes on the largest zone and 12 on the second — with hole lengths of 161
to 172 metres and inclinations of −70° throughout. Azimuths range from 130° to 250°, each set perpendicular
to its polygon’s dominant strike so that a hole intersects the modelled vein at maximum apparent thickness.

The Phase-1 confirmation subset is the number that matters commercially: one hole per zone, ten
holes, 1,925 metres of drilling, roughly 64 rig-days on a single rig. At an all-in diamond-core cost
band of USD 150 to 250 per metre — covering drilling, logging, assay and supervision — that is
USD 289,000 to USD 481,000 to put ground truth under every one of the ten satellite
anomalies. Per zone, the spread is narrow: 185 to 198 metres and roughly USD 28,000 to 49,500 each.

Need the same drill-ready geometry for your licence?
Request a project quote — or explore our
satellite-driven 3D mineral
prospectivity mapping, which delivers the interactive subsurface model and drilling-angle guidance behind
figures like the one above.

The economics — and the honest negative the model returned

At a reference gold price of USD 4,054 per troy ounce, the ten zones carry a gross in-situ value of
approximately USD 46 to 60 million. Applying the blended 81% recovery factor gives a
recoverable-metal figure of roughly 9,300 to 13,000 ounces, worth about
USD 34 to 53 million. The report’s own base case — headline metal at 81% recovery — lands at
USD 48.9 million; the wider band reflects grade uncertainty at the low end and the optimistic recovery
scenario at the high end.

The sensitivity picture is undramatic in a useful way. Across the full USD 2,850 to 5,250 band, gross
in-situ value moves from USD 42.5M to USD 78.2M and recoverable value from USD 34.4M to USD 63.3M. Every
point on that curve is positive. Gold price is not the risk on this project — grade confirmation is.

Three recovery scenarios

The honest negative: no economic pit at scoping assumptions

The delivered report also ran the detection results through the same sequence a mine-planning team would
execute in Datamine Studio, Maptek Vulcan or GEOVIA Surpac — block model, pit optimisation, grade-tonnage
curve, production schedule — explicitly labelled as a synthetic, exploration-target-grade exercise with no
drill data behind it. And the pit optimiser returned a negative result: no shell with positive value
at the registered scoping assumptions. The floating-cone optimiser found no combination of blocks
that pays for its own waste stripping.

We publish that because it is the truth, and because it reframes the project correctly. At the modelled
grades, costs and price, these mineralised envelopes do not support open-pit extraction as they stand. Any
development concept would need higher grades confirmed by drilling, a selective or smaller-scale mining
method, or materially different cost assumptions. There is a second number that makes the same point from a
different angle: the mineralised envelopes hold roughly 56.8 million tonnes of material at or above cut-off,
of which the detection model expects only about 909,000 tonnes — around 1.6% — to be recoverable
ore. Satellite anomalies outline where mineralisation sits; they do not tell you how
completely it fills the outlined volume.

Other risks on the register

• ⚠ Vein continuity at depth — MEDIUM. The continuity index averages 70.1 across the block. Phase-1 core drilling on the top-priority zones is the resolver.
• ✔ Grade variability — LOW. Polygon-level grades span a narrow 0.500 to 0.609 g/t band. A bulk metallurgical test is recommended.
• 📊 Recovery factor — LOW to MEDIUM. The 81% blended assumption is conservative for oxide gold; heap-leach response typically runs 75–90% and carbon-in-leach 88–94%.
• 📊 Gold price — LOW to MEDIUM. The sensitivity model stays positive well below USD 3,000 per ounce.
• ✔ Permitting and environment — LOW. No anomaly polygon is crossed by drainage, and the nearest building sits 74 metres from the nearest zone boundary.

Impact: months to days, and 80–85% off early-stage cost

Set the deliverable against the classical alternative. To reach the same position by conventional
means — a ranked, geometrically-specified, individually-costed target list across 283 hectares — a licence
holder would be looking at a field campaign of geological mapping, trenching and geochemical sampling
measured in months, with crews, camps, access and logistics attached. The satellite route compresses that
front end from months to days, and cuts early-exploration cost by
up to 80–85%. That is the practical case for gold anomaly detection in Zambia and across
comparably vegetated African terrain: not that it is more accurate than a drill, but that it tells you which
ten hectares of a 283-hectare block deserve one.

The turnaround on a screen like this is 5 to 20 business days from receipt of the area of interest,
depending on block size and mineral complexity. The client supplies coordinates, a KML or KMZ file, or a
drawn polygon, plus the country or region and the target minerals. We select the data source — multispectral
or hyperspectral — acquire it, run the analysis, and deliver. This is the second Zambian gold block we have
published in this form; the companion study on a smaller licence area is here:
satellite gold
exploration in Zambia — three drill-ready zones.

The phased roadmap the report recommends

1. Phase 1 (months 0–6) — confirmatory drilling. Fifteen HQ diamond core holes on the largest zone at 250° / −70°, 167 m each, plus twelve on the second zone at the same geometry. A 200 lb composite sample for bottle-roll cyanide and column-leach metallurgy. Geological mapping on a 1:1,000 grid with structural measurements at every bedrock outcrop.
2. Phase 2 (months 7–12) — expansion and resource definition. Two confirmation holes each on the next three zones. Ground magnetic and induced-polarisation survey across the whole 159-acre anomaly footprint — which is also the direct mitigation for the magnetic risk flag. Then a compliant resource estimate on a kriged block model.
3. Phase 3 (months 13–24) — pre-feasibility and permitting. Infill drilling on a 25 m × 25 m grid to reach measured-and-indicated category. Geotechnical and hydrogeological drilling, environmental baseline studies, plan-of-operations engagement with the regulator, and a pre-feasibility study with a capital estimate to ±25% accuracy.

Glossary

Gossan zone

An iron-oxide-rich, weathered surface cap formed by oxidation of an underlying sulphide body. Because it sits at surface and is spectrally distinctive, it is one of the most satellite-detectable expressions of a gold system.

Exploration target

A conceptual, range-based statement of the potential quantity and grade of mineralisation, based on indirect evidence. It is explicitly not a Mineral Resource and cannot be used for public disclosure.

Cut-off grade

The lowest grade at which material is treated as ore rather than waste. Here, 0.25 g/t Au — every zone in the study sits above it.

Strip ratio

Tonnes of waste that must be moved per tonne of ore recovered. The block averages 2.74 : 1 weighted by tonnage; surface gold operations commonly accept up to 3.0–4.0.

Dilution

Waste rock unavoidably mined with the ore, which lowers the delivered grade. An 18% allowance is applied throughout this study.

P10 / P50 / P90

Percentiles of a modelled distribution. P50 is the median; P10 and P90 bracket the central 80% of outcomes. Quoting P10–P90 rather than a single number is what makes a satellite estimate honest.

Blended recovery

Mining recovery multiplied by metallurgical recovery. The 81% base case here is 90% × 90%.

Continuity index

A measure of how coherently a mineralised structure persists along strike and at depth. The block averages 70.1, which is why depth continuity is flagged as a medium risk.

How to run gold anomaly detection on your own licence

The workflow is deliberately simple, and it does not require you to have any prior geoscience dataset.

• ✔ Send your area of interest. Coordinates, a KML or KMZ file, or a polygon drawn on a map — plus the country or region and your target mineral or minerals.
• 📊 We select and acquire the data. Multispectral or hyperspectral, chosen to suit your terrain, cover and commodity.
• ✔ You receive a premium mineral-intelligence report. High-potential zones, prospectivity heatmaps, estimated location and depth ranges, indicative quantities, geological interpretation of faults, alteration and host rock, and seasonal anomaly validation — as a PDF plus georeferenced GIS files.
• 📊 Or go Premium+ for TargetMax™ Drilling Intelligence. Optimal drilling-angle recommendations, higher ore-intersection probability, reduced drilling risk, interactive 3D subsurface models of vein structures and mineral distribution, plus commercial conclusions and next-step guidance.
• ⚠ Expect 5 to 20 business days, depending on block size and mineral complexity.

Farmonaut has applied this approach across 100,000+ hectares, 20+ mineral
types and 25+ countries spanning Africa, South America, North America, Asia and
Australia — precious metals, base metals, energy and battery minerals, industrial minerals, specialty
high-value materials and rare earth elements.

Frequently asked questions

All resource, grade and economic figures in this case study are satellite-derived exploration-target estimates reproduced from a client report, presented as ranges to reflect the underlying modelled distributions. They are not Mineral Resources or Mineral Reserves under JORC, NI 43-101, SK-1300 or SAMREC, no drilling or assaying underlies them, and they must not be relied upon for public disclosure or investment decisions. The conceptual mining study referenced is explicitly synthetic and exploration-target grade. The client’s identity, licence designation, coordinates, district and survey grid have been redacted throughout, and satellite figures have been processed to remove coordinate axes, basemap and boundary detail; delivered client imagery is full-resolution and georeferenced. Reference gold price USD 4,054 per troy ounce as stated in the source report.