Pegmatite fields in Zimbabwe and Namibia cover hundreds of square kilometres, and most of the bodies in them are barren. A pegmatite is a very coarse-grained igneous rock, and only a small share of them carry lithium minerals. You cannot walk every outcrop. Satellite data can help you cut a regional search area down to a ranked list of places worth visiting. After reading this you should be able to build a first-pass, mica-focused screening map from free imagery, say what it shows and does not show, and plan the field checks that test it.

The principle: why pegmatites are visible from space

Lithium-bearing pegmatites of the LCT family (lithium–caesium–tantalum) form from the last, most chemically evolved fractions of granitic melt. Those fractions are rich in water and in elements that normal rock-forming minerals reject. They crystallise as coarse feldspar, quartz and mica, with lithium minerals such as spodumene, petalite or lepidolite in the most fractionated bodies. From a satellite's point of view, the useful result is a surface rich in pale feldspar, quartz and white mica, set in darker or differently clay-rich host rock.

The physics is reflected sunlight. Each mineral absorbs particular wavelengths, and a sensor records how much light comes back in each band. Micas and clays that contain Al–OH bonds (aluminium bonded to hydroxyl) absorb near 2200 nm in the shortwave infrared (SWIR). One study reports that lithium-bearing rocks show absorption features at about 365, 2200 and 2350 nm. Sentinel-2's SWIR bands sit at 1.610 µm (band 11) and 2.190 µm (band 12), both at 20 m resolution. Band 12 falls inside the Al–OH absorption and band 11 sits on its shoulder, so a B11/B12 ratio rises where Al–OH minerals are abundant.

Two caveats follow. Spodumene itself is hard to identify with Sentinel-2. One review notes that the sensor lacks thermal infrared bands, which matter for silicate minerals such as spodumene and petalite. You are therefore mapping the mica-and-feldspar signature of a pegmatite, and usually not the lithium mineral. Also, a mica signature does not mean lithium. Barren pegmatites, sericitic alteration and weathered granites also carry Al–OH.

The workflow, step by step

1. Define the geological frame first. Pegmatites cluster around evolved granites and along the structures that fed them. In Zimbabwe that means greenstone belts and their granite margins. In Namibia it means the pegmatite swarms of the Damara orogen. Digitise granite contacts and shear zones from published geological maps, then use them as a spatial filter later.
2. Choose imagery. Sentinel-2 is free and has four bands at 10 m, six at 20 m and three at 60 m. ASTER offers more SWIR bands, which suits Al–OH work. Check that SWIR data exist for your dates before relying on it. Use dry-season scenes with minimal cloud and vegetation.
3. Pre-process. Use surface-reflectance products (Sentinel-2 Level-2A), resample all bands to 20 m, and build a median composite from several dry-season dates to suppress noise and cloud.
4. Mask. Remove vegetation, water, cultivated land, burn scars and settlements. Published work warns that spectral confusion with urbanised areas or agricultural fields is noticeable.
5. Compute indices. Calculate B11/B12 for Al–OH. For ASTER, a relative band depth for sericite and muscovite is (band 5 + band 7) / band 6. Add an iron-oxide ratio such as red/blue (B4/B2) to flag weathered mafic rock that could mimic or hide the signal.
6. Combine and classify. Use false-colour composites, selective principal component analysis (PCA, which compresses correlated bands into a few uncorrelated components) or a Spectral Angle Mapper (SAM, which compares each pixel's spectrum with a reference). These methods have been applied to lithium pegmatites with ASTER, Landsat and Sentinel-2.
7. Filter spatially. Keep anomalies within your granite-proximity and structural buffers, and remove pixel clusters too small to be geologically meaningful.
8. Rank and validate. Score each cluster, then compare your ranking with any known occurrences you can use as a control.

Worked example (illustrative only)

This is a hypothetical case with invented numbers. It does not describe a real project.

Suppose you hold a 400 km² block of greenstone and granite in a semi-arid setting. You build a dry-season Sentinel-2 median composite, mask vegetation and cultivation, and compute B11/B12. You then take the top 2% of masked pixels, which gives 1,800 anomalous pixels. Each 20 m pixel covers 400 m², so that is 0.72 km².

Grouping the pixels leaves 140 clusters. Dropping clusters under three pixels leaves 60, and keeping only those within 3 km of a mapped granite contact leaves 22. The iron-oxide ratio removes 5 clusters that sit on ferruginous weathered mafic rock, which leaves 17.

Resolution limits what you can see. A 30 m-wide dyke fills only about one and a half pixels, so most pixels covering it also include host rock and the signal is diluted. Ranking by cluster size alone would therefore favour wide bodies, and you would miss narrow ones. Instead, score each cluster on ratio strength, structural position and clustering. You would then send a field team to the top five or six clusters, plus two low-ranked ones as controls.

Common mistakes and limitations

• Mistaking mica for lithium. Muscovite-rich pegmatites are common, and most contain no spodumene. One mineral-systems study states that a prospective cluster does not guarantee economic spodumene. Treat the output as a prioritisation of ground, not a lithium map.
• Cover. Soil, laterite and transported sand hide the bedrock. Satellite sensors read only the top few microns of surface, so deep weathering in humid parts of Zimbabwe can erase the signal, while desert varnish and calcrete can mislead in arid Namibia.
• Cultural and seasonal noise. Mine dumps, quarries, artisanal workings and burnt ground can all produce false anomalies.
• Using one threshold everywhere. Ratios shift with illumination, soil colour and atmosphere, so tune thresholds per terrane.
• Trusting a classifier's accuracy figure. A map that scores well against its own training polygons may still fail on new ground. Hold out whole areas for testing, not just random pixels.

How to check your results

Test your index on known pegmatites first. If it fails to light up bodies you already know, it will not be reliable elsewhere. Then ground-truth in the field with a hand lens, a portable SWIR spectrometer if you have one, and sampling for laboratory lithium assay. Handheld XRF cannot measure lithium, because the element is too light, so rely on laboratory assays or handheld laser-induced breakdown spectroscopy (LIBS). Record your field results against every target, including the misses, and use them to refine the next version.

Key points to remember

• Satellites detect the mica, clay and feldspar signature of pegmatites, not lithium itself.
• Sentinel-2 bands 11 and 12 bracket the 2200 nm Al–OH absorption, so the B11/B12 ratio is a sensible starting index.
• Geological filters (granite proximity, structure) turn a spectral map into a target list.
• Masking, dry-season compositing and testing on known pegmatites matter more than the choice of algorithm.
• Resolution, cover and false positives limit what the map can show, so every ranked target needs field and laboratory confirmation.

Sources

• Constraints and Potentials of Remote Sensing Data/techniques Applied To Lithium (li)-pegmatites (researchgate.net)
• A lightweight deep learning based cloud detection method for Sentinel-2A imagery fusing multi-scale spectral and spatial features (arxiv.org)
• Detecting Lithium (Li) Mineralizations from Space: Current Research and Future Perspectives (mdpi.com)
• S2 Mission (sentiwiki.copernicus.eu)
• Potential of Sentinel-2 data in the detection of lithium (Li)-bearing pegmatites: a study case (researchgate.net)
• Towards better delineation of hydrothermal alterations via multi-sensor remote sensing and airborne geophysical data (nature.com)

About Orex — Orex is a mineral exploration intelligence platform based in Mwanza, Tanzania. We combine satellite remote sensing, elevation-derived structural analysis and open geoscience data to help explorers, licence holders and investors focus their fieldwork on the ground that matters.

Want to see fault structures and intersection targets on your area of interest — for free? Install GoldRadar on your phone or desktop: it maps lineaments and automatically flags fault intersections derived from satellite elevation data, giving you a structural framework for preliminary exploration before you spend a dollar on the ground.