Hyperspectral mineral mapping for exploration targeting
Spaceborne hyperspectral sensors resolve mineral absorption features invisible to broadband cameras, mapping alteration halos that guide drill-targeting for porphyry copper and epithermal gold systems.
Sensors
- ASI PRISMA: Hyperspectral imager covering 400–2500 nm across 237 bands at approximately 30 m pixel size. Revisit roughly 29 days at nadir, reducible by off-nadir pointing. The 30 m floor is a genuine constraint in narrow hydrothermal corridors or structurally controlled veins less than one pixel wide.
- NASA/JPL EMIT (aboard the ISS): Imaging spectrometer covering 380–2500 nm at ~60 m spatial resolution. Optimised for mineral dust source mapping but fully capable of alteration mineral discrimination. ISS orbital inclination of 51.6° limits coverage to latitudes below about 52°, excluding parts of northern Canada, Scandinavia and Russia.
- DLR/ESA EnMAP: Launched April 2022. VNIR/SWIR coverage 420–2450 nm, 30 m pixel, 30 km swath, target revisit of 27 days. Radiometric quality is high and the signal-to-noise ratio is well documented, making it competitive with PRISMA for SWIR clay and carbonate discrimination.
- Landsat 8/9 OLI-TIRS: Not hyperspectral, but the six reflective OLI bands plus two TIRS thermal bands allow coarse mineral group discrimination (clay vs iron oxide vs carbonate) through band ratios. Spatial resolution 30 m multispectral, 100 m thermal. Useful for regional screening before committing hyperspectral tasking budget.
What the absorption feature actually tells you
Every silicate, carbonate and hydroxyl-bearing mineral has diagnostic absorption features in the shortwave infrared, typically between 1300 nm and 2500 nm. Kaolinite absorbs strongly near 2200 nm. Alunite shows a doublet near 2165 nm and 2320 nm. Chlorite absorbs near 2250 nm. These are not subtle statistical artefacts; they are quantum-mechanical consequences of molecular bond vibrations, stable across illumination conditions and well-catalogued in the USGS spectral library.
The practical value for exploration is that these minerals rarely form in isolation. Alunite and advanced argillic assemblages (kaolinite, dickite, pyrophyllite) indicate high-sulphidation epithermal or the lithocap above a porphyry system. Propylitic chlorite and epidote signal the outer halo of a porphyry. Mapping their spatial distribution and zoning pattern gives a geologist a structural picture of the hydrothermal system without a single soil sample. That is the argument for hyperspectral data early in an exploration programme.
From photons to mineral map: the processing chain
Raw hyperspectral radiance must be converted to surface reflectance before any mineral identification is meaningful. Atmospheric correction is the step where most errors are introduced. Standard approaches include ACORN, FLAASH and the empirical line method; EMIT data are distributed with a pre-applied atmospheric correction based on the ISOFIT algorithm, which reduces processing burden but deserves independent validation in high-altitude or dust-laden scenes.
Once surface reflectance is in hand, mineral identification uses spectral matching against a reference library. Spectral Angle Mapper (SAM) measures the angular distance between a pixel spectrum and a reference spectrum in n-dimensional space, making it relatively insensitive to illumination magnitude. Mixture Tuned Matched Filtering (MTMF) estimates sub-pixel abundance and suppresses false positives. Neither method is infallible. Vegetation, soil moisture and shadow all distort SWIR spectra, and arid or semi-arid terrain is strongly preferred for reliable results. Heavily vegetated tropical terrains are where hyperspectral mineral mapping loses most of its discriminating power.
The output is typically a classified mineral map and a set of abundance rasters, one per target mineral group. These feed directly into a GIS alongside structural geology and geophysical data.
Resolution floors and what they cost you
At 30 m, a single PRISMA or EnMAP pixel covers 900 square metres. A narrow quartz-alunite vein cutting through a propylitic halo may be two metres wide. It will not be resolved; it will be diluted into the surrounding pixel spectrum and may not be detected at all unless it is abundant enough to shift the mixed spectrum measurably. EMIT at 60 m is coarser still.
This is not a fatal limitation for porphyry targets, where alteration halos extend hundreds of metres to kilometres and the mineral zoning is broad. It is a genuine problem for epithermal vein systems, skarn contacts and structurally controlled narrow corridors. For those targets, airborne hyperspectral systems flying at 1–5 m resolution remain the tool of choice. Spaceborne data is best used to rank and prioritise targets across a large land package before committing to airborne survey costs.
Cloud cover is a separate constraint that receives less attention than it deserves. SWIR hyperspectral sensors are passive and optical; a single cloud layer renders the acquisition useless. In tropical or maritime climates, achieving cloud-free coverage of a specific target may require multiple acquisition attempts across several months.
Landsat as a cheap first filter
Before ordering PRISMA or EnMAP scenes, a Landsat band-ratio analysis costs nothing and takes hours. The clay ratio (band 6 / band 7 in Landsat 8/9 OLI terms) highlights hydroxyl-bearing minerals. The iron oxide ratio (band 4 / band 2) maps ferric iron from gossans and oxidised sulphides. The carbonate index uses band 7 relative to bands 5 and 6. These are blunt instruments with no ability to distinguish kaolinite from alunite, but they are excellent at identifying which parts of a large land package warrant the higher cost of hyperspectral tasking.
The Landsat archive extends back to 1972 in various forms, giving decades of seasonal coverage. That depth is irrelevant to mineralogy (rocks do not change on that timescale) but useful for identifying areas with consistently low cloud cover, which is worth knowing before you plan a hyperspectral acquisition campaign.
Honest limits and what good interpretation requires
Hyperspectral mineral mapping identifies surface mineralogy. It cannot see through overburden, alluvium or vegetation. A deeply weathered laterite profile may present a uniform iron-oxide surface that tells you little about what lies beneath. Supergene enrichment and secondary mineral assemblages at surface may not reflect the primary ore system at depth. The map is a geochemical surface signal, not a resource estimate.
Interpretation requires a geologist who understands the deposit model. A pixel classified as alunite in an area with no structural control and no associated silicification is probably not a drill target. Conversely, a subtle chlorite anomaly in the right structural position, corroborated by a magnetic low and a coincident soil geochemistry anomaly, may be the most important pixel on the map. The data reduces uncertainty; it does not replace judgement.
Satellize integrates PRISMA, EnMAP and Landsat mineral mapping into GIS-ready deliverables, with spectral processing workflows built on published library matching methods. The Tonga crop-estimation programme is a different domain, but the spectral calibration discipline is the same.
Combining thermal infrared: what TIRS adds
Landsat 8/9 TIRS and, at coarser resolution, ASTER (now in archive mode) provide emissivity data in the thermal infrared between roughly 8 and 12 micrometres. Silica-rich rocks emit differently from carbonate or mafic rocks in this range, a property called the reststrahlen effect. Silicification halos around epithermal centres show a measurable emissivity contrast against surrounding andesite or dacite. At 100 m TIRS resolution the discrimination is coarse, but it is an independent line of evidence that can corroborate SWIR clay mapping and help distinguish true silicic alteration from surficial quartz sand.
ASTER's five thermal bands at 90 m resolution were more capable for this application, and the ASTER Global Emissivity Dataset remains a useful archive product even though the sensor is no longer acquiring new data routinely. For new acquisitions, TIRS is the practical option.
Typical figures
| Spatial resolution (PRISMA / EnMAP) | 30 m |
| Spatial resolution (EMIT) | ~60 m |
| Spectral range (SWIR sensors) | 400–2500 nm, 200–240 bands depending on sensor |
| Revisit (PRISMA / EnMAP) | ~27–29 days at nadir; reducible by off-nadir tasking |
| EMIT latitude coverage | Approximately 52°N to 52°S (ISS orbital constraint) |
| Minimum detectable alteration width | Reliably >1 pixel (>30 m); sub-pixel mixing reduces confidence for narrower features |
| Cloud sensitivity | Total obstruction; passive optical only, no cloud penetration |
| Archive depth (Landsat band-ratio screening) | 1972 to present (Landsat 1–9 series) |
| Thermal infrared (Landsat 9 TIRS-2) | Two bands, 10.6–11.2 µm and 11.5–12.5 µm, 100 m resolution |
| Delivery format | GeoTIFF mineral classification raster, abundance rasters per mineral group, vector anomaly polygons, GIS-ready package |
Analytics Satellize can run
| Surface mineral classification map | Spectral Angle Mapper (SAM) against USGS spectral library references | GeoTIFF classified raster with mineral group legend, delivered per scene or mosaicked across land package |
| Alteration mineral abundance rasters | Mixture Tuned Matched Filtering (MTMF) for sub-pixel abundance estimation | Per-mineral abundance GeoTIFF (e.g. alunite fraction, kaolinite fraction) with confidence layer |
| Hydrothermal zoning interpretation | Spatial overlay of mineral assemblage maps against deposit model templates (porphyry, epithermal) | Annotated PDF interpretation report with ranked target polygons and supporting spectral plots |
| Landsat band-ratio screening layer | Published OLI band-ratio indices (clay, iron oxide, carbonate) applied to full archive | Regional GIS layer flagging anomalous zones for hyperspectral follow-up tasking |
| Silicification index from thermal emissivity | Emissivity contrast mapping using Landsat TIRS or ASTER archive data | Emissivity anomaly raster overlaid with SWIR clay map for independent corroboration layer |
| Cloud-free acquisition planning report | Historical cloud-cover frequency analysis from Landsat and Sentinel-2 archive | Monthly cloud-probability map for target area, used to schedule PRISMA or EnMAP tasking windows |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.