Hyperspectral mineral alteration and lithological mapping
Spaceborne imaging spectrometers resolve diagnostic absorption features that separate kaolinite from illite, jarosite from goethite, and barren rock from altered crust. Broadband multispectral sensors cannot do this reliably. Here is what current instruments can and cannot deliver.
Sensors
- EnMAP (DLR/GFZ, launched 2022): 430 to 2450 nm range, 6.5 nm spectral sampling in VNIR and 10 nm in SWIR, 30 m spatial resolution, 30 km swath. Revisit approximately 27 days at nadir, reduced to roughly 4 days with off-nadir pointing. Designed specifically for geological and ecosystem mapping; signal-to-noise ratios published at above 400:1 in SWIR.
- PRISMA (ASI, launched 2019): 400 to 2500 nm, approximately 12 nm spectral sampling, 30 m spatial resolution, 30 km swath. Revisit nominally 29 days at nadir. Italian Space Agency instrument; data access requires registration with ASI. Atmospheric correction performance over bright arid terrain is an active research area.
- EMIT (NASA/JPL on ISS, deployed 2022): 380 to 2500 nm, 7.4 nm spectral sampling, 60 m spatial resolution, 75 km swath. Not a survey instrument: ISS orbital inclination (51.6°) limits coverage to latitudes between roughly 51.6° N and S, and the ISS ground track is non-repeat. Optimised for arid dust-source mapping; spectral library and scene products are open access via NASA Earthdata.
- ASTER SWIR (NASA/METI, legacy): Six SWIR bands between 1600 and 2430 nm, 30 m resolution. The SWIR detectors failed in 2008 due to detector saturation and cooling loss, so no new SWIR data exists. The archive (1999 to 2008) remains scientifically valuable for regional alteration mapping and is freely available via USGS EarthExplorer. Do not plan operational workflows around new acquisition.
Why 10 nm matters and 100 nm does not
Clay minerals, carbonates, sulphates and iron oxides absorb shortwave infrared energy at wavelengths governed by their crystal chemistry. Kaolinite has a characteristic doublet near 2200 nm. Illite absorbs at roughly 2210 nm. That 10 nm separation is invisible to any broadband sensor, including Sentinel-2, whose narrowest SWIR band is 20 nm wide and centred far too coarsely to resolve the doublet. Landsat 8 and 9 are no better. Multispectral sensors can flag that alteration is present; they cannot tell you which mineral is doing the altering.
Jarosite and goethite both produce broad iron-oxide signatures in the visible, but jarosite has a distinctive absorption near 430 nm and a SWIR feature near 2270 nm that goethite lacks. That distinction matters: jarosite forms in acidic, oxidising environments often associated with sulphide weathering, while goethite is far more general. Confusing them in an acid-mine-drainage survey or a porphyry copper exploration programme is not a minor error. Imaging spectrometers with 6 to 12 nm sampling resolve these features directly; multispectral sensors require indirect inference that breaks down over mixed pixels.
The three instruments doing this work now
EnMAP is the most capable operational survey instrument currently in orbit for this application. Its signal-to-noise specification, spectral sampling and 30 m ground sampling distance make it the closest equivalent to the airborne AVIRIS-NG or HyMap systems that geologists have trusted for decades. Tasking is available through DLR's ground segment; standard processing to Level 2A (atmospherically corrected surface reflectance) is part of the product chain.
PRISMA covers a similar spectral range and resolution but has a smaller user community and less published validation over complex geological terrain. It is a credible alternative for regions where EnMAP tasking is slow or unavailable, particularly across the Mediterranean, North Africa and the Middle East, where ASI has prioritised acquisitions.
EMIT was built to map mineral dust sources globally, not to prospect for ore. Its 60 m pixels and non-repeat ISS track make it unsuitable for site-scale mapping or systematic revisit. What it offers is open data, a well-validated atmospheric correction (ISOFIT-based), and coverage of many arid regions that lack EnMAP or PRISMA archive. For regional reconnaissance it is genuinely useful. For anything requiring spatial precision below 100 m, it is not the right tool.
Atmospheric correction is not a footnote over arid terrain
Bright, dry soils and exposed rock reflect strongly across the SWIR. That same brightness amplifies errors in the atmospheric path radiance estimate. Standard correction algorithms calibrated over vegetated or ocean scenes can introduce artefacts that mimic or mask real mineral absorption features. The ISOFIT algorithm, developed partly for EMIT, uses a physics-based radiative transfer model with surface and atmospheric parameters retrieved simultaneously, and performs better over bright targets than older empirical line methods.
Aerosol loading from dust events is a particular problem. A dust plume over the Sahara or the Arabian Peninsula can shift apparent surface reflectance by several percent in the VNIR, enough to corrupt iron-oxide band ratios. Scene-level quality flags in EnMAP and EMIT Level 2 products report aerosol optical depth estimates, but users should treat any scene with AOD above roughly 0.3 with caution and consider re-processing with locally measured atmospheric data where available.
What a 30 m pixel floor actually costs you
Current spaceborne imaging spectrometers share a 30 m spatial resolution floor, with EMIT coarser still at 60 m. For regional geological mapping, prospect-scale alteration halos and lithological contacts spanning hundreds of metres, this is adequate. For vein-scale features, fault gouge, or alteration pipes narrower than roughly 60 m, it is not. A 30 m pixel over a mixed outcrop of kaolinite and unaltered host rock will return a blended spectrum that sits between the two end-members. Spectral unmixing methods can recover sub-pixel abundance estimates, but they require a clean spectral library and assumptions about mixing linearity that do not always hold.
Airborne spectrometers routinely operate at 3 to 5 m resolution and remain the standard for detailed mine-site mapping. Spaceborne data is best used for regional target generation: narrowing a search area from thousands of square kilometres to tens, then directing airborne or field campaigns to the shortlisted zones. That division of labour is not a limitation to apologise for; it is an efficient workflow.
From spectra to a deliverable geological product
Raw surface reflectance cubes are not a deliverable. Turning them into a mineral map requires spectral matching against a reference library (the USGS Spectral Library, version 7, is the standard), band-ratio composites for rapid visual discrimination, and supervised or unmixing-based classification. Published methods include Spectral Angle Mapper, Matched Filter and constrained least-squares unmixing. Each has known failure modes: SAM is insensitive to brightness but sensitive to spectral shape errors; unmixing requires end-member purity that is rarely guaranteed from orbit.
The output a geologist actually needs is a GIS layer showing mineral group distributions, a confidence estimate per pixel, and a report linking spectral detections to known geological context. Satellize builds these workflows on open instrument data from EnMAP and EMIT, adding commercial tasking via client licence where archive coverage is insufficient. The Tonga crop-estimation programme is a different domain, but the underlying pipeline logic, ingesting calibrated radiance, applying physics-based correction, and extracting a thematic map, transfers directly to geological applications.
Honest caveat: spaceborne hyperspectral data identifies mineral groups, not individual ore deposits. Detecting a kaolinite-alunite assemblage consistent with high-sulphidation epithermal alteration is a meaningful exploration signal. It is not a resource estimate.
Typical figures
| Spatial resolution (EnMAP, PRISMA) | 30 m ground sampling distance |
| Spatial resolution (EMIT) | 60 m ground sampling distance |
| Spectral range | 380 to 2500 nm (VNIR + SWIR); instrument-dependent lower cutoff |
| Spectral sampling interval | 6.5 to 12 nm depending on instrument and spectral region |
| Number of bands | 224 to 250+ contiguous bands (EnMAP); ~285 bands (EMIT) |
| Revisit (nadir) | ~27 days (EnMAP); ~29 days (PRISMA); non-repeat (EMIT/ISS) |
| Revisit with off-nadir pointing | ~4 days (EnMAP at ±30° cross-track) |
| Latency to Level 2A product | Typically 1 to 5 days for EnMAP standard processing; EMIT open data updated continuously via NASA Earthdata |
| Latitude coverage (EMIT) | Approximately 51.6° N to 51.6° S (ISS inclination limit) |
| ASTER SWIR archive | 1999 to 2008 only; no new SWIR acquisition possible |
Analytics Satellize can run
| Mineral group classification map | Spectral Angle Mapper or Matched Filter against USGS Spectral Library v7; applied to atmospherically corrected Level 2A surface reflectance | GeoTIFF layer with mineral group labels and per-pixel confidence score, delivered as GIS-ready file |
| Iron-oxide and hydroxyl band-ratio composite | Published band-ratio indices (e.g. clay index, ferric iron index) computed from SWIR and VNIR bands; rapid visual discrimination of alteration zones | False-colour composite image and ratio raster stack, with interpretation notes in PDF report |
| Sub-pixel mineral abundance estimate | Constrained least-squares spectral unmixing using library end-members; abundance per mineral end-member per pixel | Multi-band abundance raster (one band per end-member), plus uncertainty layer |
| Alteration halo delineation for exploration targeting | Spatial clustering of co-occurring clay and sulphate mineral detections; geological context overlay from public mapping | Vector polygon layer of ranked target zones with spectral evidence summary, formatted for exploration GIS |
| Atmospheric correction quality assessment | Scene-level aerosol optical depth flagging and water vapour column retrieval from instrument metadata; ISOFIT or ATCOR benchmarking | Per-scene quality report flagging pixels with elevated AOD or cloud adjacency effects |
| Change detection between archive epochs | Difference of classified mineral maps across ASTER SWIR archive (pre-2008) and current EnMAP or PRISMA acquisitions; change attributed to mining, erosion or vegetation removal | Change polygon layer with date-stamped before/after spectral profiles |
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.