Part three of a three-part series on TerraEye 2.0 and the changing economics of mineral exploration.
Satellite data has been promising to revolutionise mineral exploration since the early 1970s, and plenty of geologists have the scar tissue to prove it. Ask the industry veterans why they haven’t embraced the latest wave of AI-driven remote sensing and the answer is usually some version of the same sentence: I used this data 20 years ago, and it didn’t bring any value.
Paweł Leśko, who leads sales at exploration targeting platform TerraEye, hears it constantly, and he is willing to meet the scepticism head on. So it is worth asking the question the sceptics are really asking: what can you actually learn about the ground from space? What can’t you learn? And where do the geologist and the field crew still come in?
The fingerprint in the light
The physics is the place to start, because it defines both the power and the limits of the method. Satellites capture the light reflected off the Earth’s surface, and every material reflects it differently.
“If you have two materials, they reflect different bands of light,” says Dr Wojciech Kuloza-Kaczan, TerraEye’s head of product. “We can capture the characteristics of both of them and classify: this one is this, and this one is that.”
That reflection works like a fingerprint, and modern sensors read it in far more detail than any human ever could. Where the eye sees three bands of colour, the satellites feeding TerraEye’s analysis read hundreds of bands across the spectrum, including well beyond visible light. It is why Leśko pushes back when a client tells him their geologist walked the terrain and saw no outcrops worth pursuing.
“You cannot compare the human eye to the satellite,” he says.
Read across an entire tenement, those fingerprints become maps: surface mineralogy, alteration patterns, and, in TerraEye 2.0, spectral similarity searches. A geologist can sample the spectral signature of a known occurrence, their own outcrop or a neighbour’s discovery, and search for matching ground anywhere in the world. Draped over the platform’s new 3D terrain, the results let a geologist read mineral accumulations against topography and structure, tracing where material may have travelled from.
That is a serious amount of information for a dataset that requires no permits, no mobilisation, and no field season. But it comes with hard boundaries, and TerraEye is blunt about them.
What the satellite cannot tell you
“There was a time people were asking directly: you’re using AI, so tell me, where is the gold?” Leśko says. “Our technology, unfortunately, is not providing information on where the gold is. We can show which part of your parcel seems to have the value, where to invest your money.”
The honesty extends to the chemistry. Some minerals simply do not produce a distinctive spectral response, and TerraEye’s geologists say so. Leśko describes a meeting with one of the world’s largest mining companies, whose exploration lead asked for a specific mineral. The answer from TerraEye’s team was that the satellite cannot identify that mineral directly, but it can detect a combination of other minerals that, in the geological concept, typically travel with it. The client accepted the logic instantly, because it is the same proxy reasoning geologists apply in the field every day.
Two more limits matter. The first is depth: spectral data reads the surface, full stop. “If you would like to go further with understanding what’s underneath the surface, then you can go with geophysics, and it gives you an approximation of the parameters of the ground below,” Kuloza-Kaczan says, which is exactly why TerraEye 2.0 now folds magnetometry, gravimetry, and geochemical datasets into the same workspace. The second is cover. “The biggest limitation is vegetation cover,” Leśko says. “It works pretty well in well-exposed terrain.” Deserts, arid shields, and sparsely vegetated ground are ideal; jungle is not. It is an admission that conveniently maps onto much of Australia, the Middle East, North Africa, and the Andes, but an admission all the same.
How reliable are the results?
Reliability, in TerraEye’s telling, is less about the algorithm and more about the workflow around it. The platform deliberately does not auto-generate drill targets, because its makers know exactly how a black box lands with this audience. Geologists, Kuloza-Kaczan says, want to understand the layers, not accept an oracle’s verdict.
Users can click any pixel and inspect the raw spectrum behind a classification, compare it against mineral libraries, and follow step-by-step interpretation guidance drawn from the company’s consulting practice. Results are designed to be tested against other evidence, not taken on faith: aggregated with national datasets, cross-validated against geophysics, and checked against whatever ground truth exists.
For the sceptics demanding proof rather than physics, Leśko points to the record. “People were asking for the proof,” he says. “That’s why we built the case studies, to show the traction between the satellite data and the work that should be done on the ground.”
Where the geologist still comes in
Everywhere, is the short answer. The satellite produces evidence; the human produces the interpretation, the targets, and the decision.
“We keep the human in the loop in the interpretation phase,” Kuloza-Kaczan says. The platform generates spectral similarity results and mineral layers; the geologist reads the hotspots, weighs them against the geological setting, and draws the targets by hand.
And nothing in the workflow ends at the screen. The best practitioners treat satellite results as a first pass that earns, or rules out, the next field dollar. Leśko cites a client who ran TerraEye over his ground as a first step and, on the strength of the prediction, chose the moment to commission geophysics. In the greenfield extreme, the field validation can be startlingly precise. The major that asked about the proxy minerals defined success as a single confirmed pixel: one verified occurrence, on ground with no historical data and no artisanal workings, would be enough to send a field team straight to the spot.
That is the frontier case, and it is where remote sensing stops being optional altogether. Mineral exploration rests on four pillars: remote sensing, geochemistry, geophysics, and structural analysis. On ground that has never been systematically explored, whether frontier belts in Africa and the Middle East or the under-mapped terrain that draws ASX explorers to places like Papua New Guinea, three of those pillars simply do not exist yet.
“When three of the four pillars are not accessible, which means we are going to a country where no exploration was done for decades, no geophysics, no geochemistry, then remote sensing is the only way to actually get knowledge about extensive areas,” Kuloza-Kaczan says.
Familiar data, new perspective
For Australian explorers, these are not abstract questions. Australia may be the world’s most spectrally literate exploration market: nationwide surveys built on ASTER data, a dataset TerraEye itself draws on as a supplier, low vegetation cover, and geologists who have worked with this kind of data for decades.
“Australia is really smart regarding spectral data,” Kuloza-Kaczan says. “There is good ground for us to check if someone would like to see a new perspective on the same kind of datasets they’ve worked with in the past.” The industry will have the chance to put its questions to the company directly when TerraEye makes its Australian debut at IMARC in Sydney this October (27 to 29).
The honest answer to “what can a satellite tell you about your ground?” is: more than the sceptics think, less than the hype merchants claim, and nothing that replaces the geologist who decides what it means. TerraEye’s bet is that an industry that distrusts magic boxes will trust an instrument that shows its working, and admits what it cannot see.
