What is a false color composite?

A false color composite assigns non-visible bands (such as near-infrared or shortwave infrared) to the red, green, and blue display channels. Because the eye sees combinations of bands it could not see directly, materials that look identical in true color become separable.

Which Sentinel-2 bands are best for geology?

For lithological discrimination a common composite is SWIR-NIR-Red as RGB, using bands 12, 8, and 4. Band 11 and 12 (SWIR) respond to clay and carbonate alteration, while the red-edge and red bands (4, 5) plus band 8 help separate iron-bearing surfaces and vegetation.

Why use ASTER for mineral mapping?

ASTER has six shortwave-infrared bands between roughly 1.6 and 2.43 micrometres, the wavelength region where clays, carbonates, and sulphate alteration minerals have diagnostic absorption features that broadband sensors like Landsat cannot resolve.

Is a false color composite the same as a band ratio?

No. A composite displays three bands as RGB to aid visual interpretation. A band ratio divides one band by another to suppress topographic shading and enhance a specific spectral feature, then can be combined into a ratio composite for targeting.

False Color Composites for Geology Mapping

Short answer

A false color composite assigns infrared and other non-visible bands to the red, green, and blue display channels so that rock types, soils, and alteration minerals that look identical to the eye become visually distinct. For geology the most productive composites use the shortwave infrared (SWIR) region, where clays, carbonates, and iron-related minerals have diagnostic spectral behaviour. The exact bands differ by sensor — Sentinel-2 bands 12/8/4, Landsat 8/9 OLI bands 7/5/2, and ASTER's six SWIR bands all serve different scales of work — and a composite is a screening and interpretation tool, not a substitute for ground checking.

Why false color, not true color

Rock-forming and alteration minerals are spectrally distinctive mostly outside the visible range. Iron oxides (hematite, goethite, jarosite) absorb in the blue and near-infrared and reflect strongly in the red; hydroxyl-bearing clays (kaolinite, illite, montmorillonite) and carbonates absorb sharply in the SWIR around 2.2 and 2.3 µm. A true-colour RGB throws away all of this. By routing infrared bands into the display, a false color composite makes those absorption and reflectance patterns visible as colour differences a mapper can interpret.

A composite differs from a band ratio. The composite is for visual interpretation across a whole scene; a ratio (e.g. SWIR1/SWIR2) divides two bands to cancel illumination and isolate one mineral's absorption, and is better for targeting. The two are complementary: map structure and lithology with composites, refine targets with ratios.

Sensor choice drives everything

Sentinel-2 (Copernicus, 10–60 m, ~5-day revisit). Thirteen bands, free, frequent. The 10 m visible/NIR bands (2, 3, 4, 8) give detail; the 20 m red-edge (5, 6, 7) and SWIR (11 ≈ 1.61 µm, 12 ≈ 2.19 µm) bands carry most of the geological information. Best general-purpose choice for regional mapping.

Landsat 8/9 OLI (USGS, 30 m, 16-day revisit). Long heritage back to the 1980s (with earlier sensors), so it is the standard for change and historical context. SWIR bands 6 (≈1.6 µm) and 7 (≈2.2 µm) underpin most Landsat alteration work.

ASTER (NASA/METI, 15/30/90 m). Its decisive advantage is six SWIR bands (bands 4–9, ≈1.6–2.43 µm) that resolve the diagnostic absorption features of specific clays, carbonates, and sulphates that Landsat and Sentinel-2 lump together. ASTER SWIR detectors degraded after 2008, so check acquisition date. Five thermal-infrared bands additionally help map silica and quartz-rich rocks.

Band combinations that work

These are written as RGB = (red channel, green channel, blue channel).

Sentinel-2 lithology / general geology: RGB = (B12, B8, B4) — SWIR2, NIR, Red. Vegetation appears dark/green-suppressed, bare rock and soils separate by SWIR brightness, and iron-rich surfaces stand out in the red channel. A second useful one is RGB = (B11, B8, B2).

Landsat 8/9 geology classic: RGB = (B7, B5, B2) — SWIR2, NIR, Blue. The long-standing "geology" composite; SWIR2 in red emphasises clay-altered ground.

Iron-oxide emphasis: push the red band high. For Sentinel-2, the iron-oxide ratio B4/B2 (red/blue) brightens hematite/goethite-rich surfaces; load it into the red channel of a ratio composite.

Clay / hydroxyl alteration: the SWIR ratio B11/B12 (Sentinel-2) or B6/B7 (Landsat) highlights Al-OH clay absorption near 2.2 µm. ASTER's classic clay index is band 5/band 6 region; the well-known ASTER ratio composite RGB = (4/6, 4/7, 4/9) separates alteration mineral groups.

Always validate the actual band-to-wavelength mapping in the product metadata before trusting any tutorial's band numbers — they differ between providers and even between Sentinel-2 processing levels.

Building a composite, step by step

Work from Level-2A / surface reflectance products (atmospherically corrected) whenever the output supports decisions. For Sentinel-2 that is the L2A product; for Landsat, Collection 2 Level-2 surface reflectance.

In GDAL, stack the chosen bands into a 3-band VRT/GeoTIFF in the RGB order you want:

gdalbuildvrt -separate -o s2_geo.vrt \
  T33TWN_B12.jp2 T33TWN_B08.jp2 T33TWN_B04.jp2
gdal_translate s2_geo.vrt s2_geo_comp.tif -co COMPRESS=DEFLATE

Sentinel-2 SWIR bands are 20 m and NIR/Red are 10 m, so resample to a common grid first or gdalbuildvrt will refuse to align them:

gdalwarp -tr 20 20 -r bilinear T33TWN_B08.jp2 B08_20m.tif

Then apply a per-band contrast stretch (2%–98% percentile clip) so the absorption differences are visible — without a stretch, raw reflectance composites look flat and grey. In QGIS, load the three bands as a single multiband raster, set Render type to Multiband color, assign bands to R/G/B, and choose Cumulative count cut 2–98%. Reproject to your project's UTM zone with gdalwarp -t_srs EPSG:326NN so the composite overlays your other layers.

Worked example: alteration reconnaissance

For a porphyry-style target in an arid, low-vegetation belt:

Acquire a low-cloud Sentinel-2 L2A scene (and an ASTER L2 scene if available).
Build RGB = (B12, B8, B4) for the structural/lithological overview and a hillshade-draped version for context.
Compute the iron-oxide ratio (B4/B2) and Al-OH clay ratio (B11/B12); make a ratio composite RGB = (B4/B2, B11/B12, B6 red-edge) to co-render iron and clay alteration.
If ASTER is available, add RGB = (4/6, 4/7, 4/9) to discriminate kaolinite vs. muscovite vs. carbonate groups.
Overlay mapped faults and known occurrences, then mark candidate zones where iron and clay signatures coincide near structures — and flag them for field/spectrometer follow-up, not as confirmed alteration.

Common pitfalls and why they happen

Mixing dates without correction. Two scenes from different seasons differ in sun angle, atmosphere, and soil moisture, so apparent "colour change" is illumination, not geology. Use surface-reflectance products and compare like dates.
Vegetation masquerading as mineralogy. Green cover dominates NIR and can mimic or hide alteration. Mask vegetation with an NDVI threshold (e.g. NDVI > 0.3) before interpreting.
Reading colour as lithology. A composite shows spectral contrast, not rock identity. The same colour can be two different materials; always cross-check with geology and field data.
Skipping the contrast stretch, or stretching the whole composite jointly. Stretch each band independently or the diagnostic differences wash out.
Ignoring spatial resolution. A 30 m Landsat pixel averages everything within it; small alteration footprints below pixel scale will not appear.

Reading the result: what the colours mean

Interpretation is where composites earn their keep, and it depends entirely on which bands you routed where. In a Sentinel-2 RGB = (B12, B8, B4) composite over arid terrain, healthy vegetation appears in muted greens because chlorophyll reflects strongly in the NIR (B8, green channel) while absorbing in red; clay-altered and weathered surfaces tend toward orange-brown because they reflect in SWIR2 (B12, red channel) and red; fresh, unaltered mafic rock often reads darker and bluer. Water is near-black across NIR and SWIR. None of these are diagnostic on their own — soil moisture, particle size, surface coatings, and shadow all shift the colour — which is exactly why a composite is a hypothesis generator, not a conclusion.

The discipline that separates useful interpretation from wishful thinking is anchoring to the known. Find a location in the scene where you already know the lithology (a mapped outcrop, a prior sample site) and learn what colour it produces in your specific composite and stretch. Then extrapolate that colour-to-material association outward, and treat every extrapolation as a target to verify rather than a fact. A composite built from a different scene, date, or stretch will not reproduce the same colours, so the association is local to that product.

Quality checks

Confirm each band's central wavelength from the product metadata, not from memory.
Verify the scene is surface reflectance (L2A / Collection 2 L2), with cloud and cloud-shadow masks applied.
Spot-check colours against any mapped lithology or known occurrence in the scene.
State sensor, processing level, acquisition date, band-to-channel mapping, and stretch in the figure caption so the composite is reproducible.

Bathyl perspective

We use false color and ratio composites as evidence layers that point a geologist toward ground worth visiting, never as automatic mineral maps. Every composite we deliver carries its sensor, date, exact band mapping, and the masks applied, so the next analyst knows precisely what the colours represent and what still needs verification.

False Color Composites for Geological Mapping