Surface Mineralogy Mapping: What Spectral Libraries Actually Tell Us

Last Tuesday I was looking at a Sentinel-2 scene over Chilas, and a pixel that looked like plain gray rock to my eye was screaming chlorite-epidote-sericite at the algorithm. That single 20-meter pixel mapped to a classic propylitic alteration halo — the kind that wraps around porphyry copper systems.

That's the whole magic of spectral libraries. And also the whole trap.

What a spectral library actually is

A spectral library is just a giant reference book of how minerals reflect light. Every mineral on Earth has a fingerprint — specific wavelengths it absorbs, specific ones it reflects. Kaolinite has a sharp dip at 2.20 micrometers. Alunite at 2.17. Muscovite around 2.20 but with a different shoulder. Chlorite drops at 2.25 and 2.33.

The USGS spectral library is the granddaddy of these. It's free, public, and contains lab measurements of over 2,400 samples — minerals, rocks, vegetation, even man-made stuff. ECOSTRESS (formerly the JPL library) is the other big one. Between them you've got most of what a working exploration geologist needs.

Here's the thing nobody tells you when you start — the libraries were built in labs, with pure samples, under perfect lighting. The real world is dusty, weathered, partly covered in dry grass, and lit by a sun that's never directly overhead. So a spectral library mineral match in lab conditions and a spectral library mineral match on a Sentinel-2 pixel over Khuzdar are two different conversations.

How we actually use them at GeoMine AI

When we run mineral spectra analysis on a target area, we're doing something called spectral unmixing. Each satellite pixel is a mixture — maybe 40% quartz, 20% iron oxide, 15% clay, 25% dry vegetation. The algorithm compares that mixed signature against library references and tries to figure out the recipe.

For Sentinel-2 we get 13 bands, which is enough to flag broad categories — iron oxides (hematite, goethite, jarosite), clay groups (kaolinite, illite, montmorillonite), and carbonates. For ASTER, with its 14 bands including shortwave infrared, we can push further into specific alteration minerals. That's where the porphyry copper hunting gets serious. ASTER's bands 5-9 are basically designed to discriminate the exact clays you see around copper systems.

Look, I got this wrong at first. Early on with breeze geo mineral analysis runs over Chagai, I was treating every kaolinite hit as a potential alteration zone. Turns out a lot of those pixels were just weathered granite. Kaolinite forms from rain too, not just from rising acidic fluids cooking the host rock. The library tells you what's there. It doesn't tell you why it's there.

That distinction is everything in exploration.

What the libraries miss (and what that costs you)

Three things, mostly.

First — gold. There's no spectral signature for gold at the concentrations we care about. Nobody is mapping ppm-level gold from space. What we're actually mapping is the alteration footprint around gold systems: silica caps, advanced argillic clays, iron oxide gossans. The library gives you the wrapper, not the candy inside.

Second — vegetation. A pixel that's 60% dry scrub and 40% rock will give you a signature dominated by cellulose and lignin absorptions, not minerals. In parts of Gilgit Baltistan this isn't a huge problem (47% of my own lease areas have less than 10% vegetation cover, which is why we love them). But try this in Azad Kashmir or KP forest belts and you're fighting the chlorophyll signal the whole time.

Third — depth. Surface mineralogy is surface. A buried orebody 30 meters down doesn't show up in any spectral library match unless something's leaked to the surface — a gossan, a soil geochemistry halo, a vegetation stress signature. This is why we pair the geo mining spectral work with SAR for structure and DEM for drainage anomalies. One data source alone lies to you. Three sources together tell something closer to the truth.

The Pakistan-specific stuff

Pakistan's geology is genuinely lucky for spectral mapping. Most of Balochistan, Chagai, Waziristan, and large chunks of Gilgit Baltistan are sparsely vegetated. Outcrop is exposed. The sun angles are workable for most of the year. Compare that to trying to map alteration through the canopy in Congo or Indonesia — we're playing on easy mode and the country has barely started using it.

The Reko Diq district, for what it's worth, shows up beautifully in ASTER alteration maps. The classic phyllic-argillic-propylitic zoning is right there in the spectral signature, just like the textbooks describe for porphyry copper systems. Anyone with a laptop and free satellite data could've flagged that ground twenty years before the drilling confirmed it. Some did. Most didn't.

That gap — between what the data was already saying and what people bothered to look at — is most of why geomines.org exists. Pakistan has roughly $6 trillion in mineral reserves sitting under ground that's been mapped from orbit for two decades. The spectral library mineral matches were there. The compute was cheap. What was missing was somebody putting it together for the people who actually own the leases.

So when someone asks me what spectral libraries tell us, my honest answer is: they tell us where to look harder. Not where to drill. Where to walk, where to sample, where to spend the next $50,000 of your exploration budget instead of the next $5 million. That's a useful answer. It's just not the answer most people want when they're hoping satellites will hand them a treasure map.

Is the rock at 36.2°N, 74.8°E really hosting a sericite-pyrite alteration halo, or is it just weathered biotite gneiss catching the afternoon light wrong? The library has an opinion. You still have to go check.