Which satellite bands are useful for geological mapping?

The visible-to-NIR bands help with iron oxides, while the shortwave infrared (SWIR, roughly 1.5-2.5 micrometres) is the key region for clays, micas and carbonates because their diagnostic absorption features sit there. Sentinel-2 and Landsat have broad SWIR bands; ASTER has six narrower SWIR bands that resolve more mineral detail.

What band ratios detect iron oxides, clays and carbonates?

For iron oxides a red/blue ratio (Sentinel-2 B4/B2, Landsat 8 B4/B2) highlights ferric iron. For clays and hydroxyl-bearing minerals a SWIR ratio such as Landsat 8 B6/B7 or ASTER B4/B6 works. Carbonates respond to ASTER B13/B14 in the thermal region. Ratios suppress topographic shading and isolate spectral slope.

Is Sentinel-2, Landsat or ASTER better for mineral mapping?

ASTER's six SWIR bands make it the classic choice for alteration-mineral mapping, though its SWIR sensor failed in 2008 so only archive data exists. Sentinel-2 offers 10-20 m resolution and frequent revisit but only two broad SWIR bands. Landsat gives the longest archive for change context. Most projects combine them.

How do I validate a multispectral geology interpretation?

Cross-check anomalies against existing geological maps, field samples or spectral library reference spectra, mask vegetation and shadow first, and confirm the same anomaly persists across dates and is not an artefact of one scene's atmosphere or sun angle.

Multispectral Imagery for Geology

Short answer

Multispectral imagery is useful in geology because rocks and alteration minerals have diagnostic reflectance and absorption features at specific wavelengths, and band ratios or composites can map those features over large, inaccessible areas. The practical skill is matching the mineral to the sensor: iron oxides show up in the visible/near-infrared, while clays, micas, and carbonates are diagnosed in the shortwave infrared (SWIR), which is why ASTER's six SWIR bands and Landsat's two SWIR bands matter far more for lithology than the pretty true-colour image. None of it replaces field validation — multispectral output is an evidence layer, not a geological map.

The physics that makes this work

Reflectance spectroscopy underpins all of geological remote sensing. A mineral absorbs light at wavelengths set by its electronic transitions and molecular vibrations, leaving dips ("absorption features") in its reflectance spectrum. Three families dominate practical mapping:

Iron oxides/hydroxides (hematite, goethite, jarosite) absorb in the blue and near the 0.85-0.9 micrometre region from ferric iron electronic transitions, so they appear bright in the red relative to the blue.
Clays, micas, and other hydroxyl-bearing minerals (kaolinite, illite, sericite, chlorite) have Al-OH and Mg-OH absorption features around 2.2 and 2.3 micrometres in the SWIR.
Carbonates (calcite, dolomite) absorb near 2.3-2.35 micrometres and have a feature in the thermal infrared near 11 micrometres.

The consequence for sensor choice is direct: a true-colour composite tells you almost nothing about lithology, because the diagnostic features live in the NIR and SWIR. Mapping mineralogy means working in the bands most people ignore.

Matching the sensor to the question

Sentinel-2 (Copernicus) carries 13 bands at 10, 20, and 60 m. It has strong visible/red-edge/NIR coverage and two broad SWIR bands (B11 ~1.61 micrometres, B12 ~2.19 micrometres) at 20 m. Its five-day revisit and free access make it excellent for iron-oxide work, regional reconnaissance, and multi-date analysis, but two broad SWIR bands cannot separate individual clay minerals.

Landsat 8/9 OLI gives 30 m multispectral data and the longest continuous archive (back to 1972 across the program), which is unmatched for historical change context. It has two SWIR bands (B6 ~1.6 micrometres, B7 ~2.2 micrometres) useful for a clay/hydroxyl ratio.

ASTER is the classic alteration-mapping instrument because it has six narrower SWIR bands plus five thermal bands. That SWIR detail lets it distinguish kaolinite from illite from chlorite, and the thermal bands let it map carbonates and silica. The critical caveat: ASTER's SWIR detector failed in April 2008, so SWIR data exists only as archive acquired before that date. Plan accordingly — for current SWIR you are on Sentinel-2 or Landsat, with ASTER as a high-detail historical reference.

A common operational mix is Sentinel-2 for current iron-oxide and broad-clay screening, ASTER archive for detailed alteration mineralogy where pre-2008 coverage exists, and Landsat for the long time series.

Band ratios and composites that earn their place

Band ratios divide one band by another. They suppress topographic shading (illumination affects both bands roughly equally and cancels in the ratio) and isolate the spectral slope that signals a mineral. The well-established ones:

Iron oxides: red/blue. Sentinel-2 B4/B2; Landsat 8 B4/B2. High values flag ferric iron staining and gossans.
Clays / hydroxyl minerals: a SWIR ratio bracketing the 2.2 micrometre Al-OH feature. Landsat 8 B6/B7; ASTER B4/B6.
Ferrous iron: a SWIR/NIR ratio (Landsat 8 B6/B5).
Carbonates (ASTER thermal): B13/B14.

A widely used display is the Sabins / Abrams-style false-colour composite: load three ratios into the R, G, B channels so different alteration types separate by hue. The point is not the colours but that each channel encodes a physical absorption feature.

In QGIS, the Raster Calculator computes a ratio directly ("B4@1" / "B2@1"); guard against division by zero and stretch the output with a 2-98 percent clip. With GDAL the same is a one-liner:

gdal_calc.py -A B4.tif -B B2.tif --calc="A.astype(float)/B" \
  --NoDataValue=0 --outfile=iron_oxide_ratio.tif

A worked reconnaissance workflow

Target: screen a semi-arid licence block for hydrothermal alteration using Sentinel-2.

Acquire a Level-2A (surface reflectance, atmospherically corrected) tile with low cloud over the dry season. Atmospheric correction matters — uncorrected top-of-atmosphere reflectance distorts the very absorption depths you are measuring.
Reproject and clip to the project CRS, e.g. gdalwarp -t_srs EPSG:32633 -cutline aoi.geojson -crop_to_cutline in.tif out.tif for a UTM zone 33N block.
Mask confounders. Build a vegetation mask (NDVI from B8/B4 above ~0.3), a water mask, and a cloud/shadow mask from the scene classification layer. Vegetation and its NIR/SWIR signature will otherwise masquerade as mineralogy.
Compute ratios: B4/B2 for iron oxide, B11/B12 as a broad clay proxy.
Composite the ratios into an RGB and apply a percentile stretch.
Interpret against context: overlay the existing geological map, structural lineaments, and any known occurrences. Alteration that aligns with mapped faults or intrusive contacts is far more credible than an isolated bright pixel.
Validate with field samples or, ideally, a portable spectrometer, and check the anomaly persists in a second-date scene.

QA and validation

Confirm atmospheric correction (Level-2A / surface reflectance) before trusting any absorption-based ratio.
Mask vegetation, water, shadow, and cloud first — most "false" mineral anomalies are unmasked confounders.
Reference against spectral libraries. Compare the scene's behaviour to known mineral spectra (e.g. the USGS Spectral Library) rather than guessing from colour.
Require multi-date persistence. A real alteration zone is stable across acquisitions; a one-scene artefact is not.
Ground-truth the anomalies that drive decisions, even with a handful of samples.

Common pitfalls and why they happen

Reading false colour as lithology. The diagnostic information is in NIR/SWIR ratios, not RGB hue; colour alone is not evidence.
Ratioing top-of-atmosphere data. Atmosphere alters absorption depth, so ratios on uncorrected data measure the air as much as the rock.
Expecting clay discrimination from two broad SWIR bands. Sentinel-2 and Landsat cannot separate individual clays the way ASTER's six SWIR bands can; do not over-interpret.
Comparing scenes across seasons without checking sun angle and moisture. Soil moisture and illumination change reflectance and create spurious "change."
Ignoring vegetation. Green cover hides and mimics mineral signatures; an unmasked NIR signal is the most common source of false positives.

Bathyl perspective

We treat multispectral products as targeting layers that point fieldwork where it is most likely to pay off, never as a finished geological map. Each anomaly we deliver states the sensor, the bands used, the masks applied, and the validation status, so a geologist can weigh it against their own observations rather than inheriting a coloured guess.