Key Takeaways
- Modelling alteration proxies has shifted from “a geologist draws domains from logging” to “hyperspectral data informs the domains directly.”
- Leapfrog Geo’s implicit modelling engine (FastRBF) is the standard tool for building the 3D alteration surfaces themselves.
- Corescan’s hyperspectral alteration maps are increasingly the data source feeding those surfaces, rather than logged intensity scores alone.
- This is still “Emerging.” The hyperspectral-to-alteration-domain pipeline is not a single push-button product yet. It is an integration you build.
TL;DR
To model alteration proxies with AI, scan your core with a hyperspectral imager and let software classify the alteration minerals along every metre. Then feed that dense, objective record into Leapfrog Geo’s implicit modelling engine alongside the geologist’s logging. As a result, domain boundaries rest on measured mineralogy rather than visual judgement alone.
What Are Alteration Proxies?
Hydrothermal fluids change the minerals in rock around an ore deposit. That change is called alteration. Typical products include sericite, chlorite, kaolinite and other clays, and each one forms under different temperature and fluid conditions. Alteration therefore acts as a proxy: it points toward the ore and tells you how the rock will behave.
For geometallurgy, alteration matters twice. First, it helps geologists vector toward mineralisation. Second, it predicts processing behaviour, because clays and micas affect grinding, flotation and tailings. An alteration model is the 3D version of that information: surfaces and domains that show where each alteration type sits in the deposit.
How Do I Predict Alteration Proxies With AI?
The workflow has three stages: measure, interpolate, then check the result against the geology you know.
Measure alteration proxies with hyperspectral scanning
Traditionally, a geologist logs alteration intensity and type at the core by eye. That logging is useful, but it is subjective. Two geologists can score the same interval differently, especially in transitional zones.
Hyperspectral core-scanning systems such as Corescan’s HCI produce a continuous, objective alteration-mineral map along the entire core. They detect clay content, sericite, chlorite and other indicator minerals spectroscopically rather than visually.
Build 3D surfaces in an implicit modelling engine
Next, the point data goes into an implicit modelling package. Leapfrog Geo dominates this step. Its FastRBF engine builds alteration, grade and fault surfaces directly from point and structural data. You do not need the manual cross-sections that older explicit workflows required. The same approach underpins AI-assisted implicit geological modelling of lithology and structure.
Combine scanned and logged alteration proxies
Finally, feed the hyperspectral dataset in place of, or alongside, logged intensity scores. In practice, this tends to produce more consistent domain boundaries. The gain is largest where visual logging is genuinely ambiguous. However, keep the geologist in the loop. Spectral classes still need geological interpretation before they become meaningful domains.
What to Expect in Practice
This is real practice at operations that have invested in core-scanning infrastructure. Even so, it is not a turnkey pipeline. You still build the integration between the Corescan output format and your geological modelling package. No single vendor sells “hyperspectral-to-alteration-model” as one product. Therefore, budget for that integration work rather than expecting to buy it off a shelf.
The alteration model also connects to the rest of the geomet chain. Clay-rich domains feed directly into clay mineral prediction for the block model. Downstream, those domains help explain differences in recovery and throughput at the plant.
FAQs
Why are alteration proxies important in mining?
Alteration proxies show where hydrothermal fluids have changed the rock, which often points toward ore. They also predict processing behaviour, because clays and micas affect grinding, flotation and tailings. A good alteration model therefore helps both exploration and plant planning.
Can hyperspectral data replace geological logging of alteration proxies?
Not entirely. Hyperspectral scanning gives a denser and more objective mineral record than visual logging. However, a geologist still interprets what the spectral classes mean for the deposit. Most sites use both together.
What software builds 3D alteration models?
Leapfrog Geo is the most widely used implicit modelling package for alteration surfaces. It interpolates logged or scanned point data into 3D surfaces with its FastRBF engine. Other implicit modelling packages can do the same job.
How mature is AI modelling of alteration?
It is emerging. The scanning hardware and the modelling software are both production-grade. The link between them, however, is still an integration each site builds for itself.
Try It With Geocluster
Working out how to wire hyperspectral scan output into your implicit modelling workflow, and what other operations have actually done, is a research question Geocluster can help you answer. It pulls real case studies rather than vendor marketing. Read how the Geocluster Research Harness works to see the approach.