Can Sentinel-2 be used for geological mapping?

Yes, for reconnaissance-scale lithological discrimination, structural lineament mapping and broad alteration screening. Its two SWIR bands plus 10 m visible and NIR bands let you separate iron-stained, clay-altered and fresh ground, but field validation is still required.

What can Sentinel-2 detect that Landsat cannot?

Sentinel-2 offers 10 m resolution in the visible and NIR (versus 30 m for Landsat) and a 5-day revisit, so structural detail, drainage textures and small alteration patches are sharper. Landsat still wins on archive length and thermal bands.

What are the limits of Sentinel-2 for mineral exploration?

It has only two broad SWIR bands, so it cannot identify specific alteration minerals the way ASTER's six SWIR channels or hyperspectral sensors can. It also lacks a thermal band, and vegetation or cloud cover can hide the surface entirely.

How do I map faults and lineaments with Sentinel-2?

Apply directional edge filters or hillshade-style enhancement to a single 10 m band (often B8), combine with a DEM hillshade from multiple sun azimuths, and trace linear features that persist across illumination directions to avoid illumination bias.

Sentinel-2 for Geological Interpretation

The short answer

Sentinel-2 is a strong reconnaissance instrument for geological interpretation: at 10 m in the visible and near-infrared and 20 m in the shortwave infrared, with a 5-day revisit from the twin satellites, it supports lithological discrimination, structural and lineament mapping, drainage analysis and broad-scale alteration screening. What it cannot do is identify specific minerals, because it has only two wide SWIR bands and no thermal channel. Treat it as the layer that tells you where to look in detail, not as a substitute for ASTER, hyperspectral data or fieldwork.

What "geological interpretation" means here

Interpretation from optical imagery rests on three things the sensor can actually resolve:

Spectral contrast between materials (iron oxides, hydroxyl minerals, carbonates, vegetation, water).
Spatial texture and pattern — drainage networks, bedding traces, fracture sets, talus, dune fields.
Structural geometry — lineaments, offsets, fold traces, lithological contacts.

Sentinel-2 contributes to all three, but its strength is texture and structure at 10 m plus coarse spectral alteration screening. The interpreter combines these with topography (a DEM) and existing geological maps.

Lithological discrimination

Most lithological separation on Sentinel-2 comes from the SWIR and from iron in the visible. The practical toolkit:

SWIR false-colour 12-11-2 separates clay-altered, carbonate-rich and fresh ground (see the companion article on band combinations for details).
Ferric iron ratio B4/B2 highlights oxidised, weathered and gossanous surfaces.
Al-OH / clay ratio B11/B12 rises over argillic alteration and clay-rich sediments.
Carbonate sensitivity is weak on Sentinel-2 because the diagnostic carbonate absorption sits near 2330 nm, at the edge of B12's response. Do not expect reliable limestone/dolomite separation; this is a job for ASTER.

A useful principal-component approach is the Crósta technique (feature-oriented PCA): run PCA on a carefully chosen four-band subset (for example B2, B4, B8, B11 for iron, or B2, B4, B11, B12 for hydroxyl) and inspect the component whose eigenvector loadings have the right sign pattern for the target mineral group. The "right" component changes per scene, so always read the eigenvector matrix rather than assuming PC2 is always the alteration component.

Structural and lineament mapping

Sentinel-2's 10 m bands are excellent for tracing faults, fractures, dykes and lithological contacts. The workflow:

Use a single high-resolution band (B8 or B4) and apply directional edge enhancement. In QGIS, the r.mfilter or a custom convolution kernel works; GDAL users can apply Sobel-like filters via gdal_calc.py or scripting.
Combine with DEM hillshades generated from several sun azimuths (e.g. 045, 135, 315 degrees) using gdaldem hillshade -az. Lineaments that appear under only one illumination direction are usually illumination artefacts; real structures persist across azimuths.
Map lineaments at a consistent scale and record azimuths so you can build a rose diagram and compare fracture orientation with known regional stress directions.

The 5-day revisit also lets you pick scenes with low sun angle (winter or early morning equivalent), which exaggerates subtle relief and makes structural features pop.

A practical end-to-end workflow

Acquire L2A for your area, choosing scenes under ~10% cloud and consistent season. Use the Copernicus Data Space or an Earth Engine ImageCollection (COPERNICUS/S2_SR_HARMONIZED).
Mask cloud and shadow with the SCL band (drop classes 3, 8, 9, 10) or the s2cloudless probability mask. For multi-date work, build a median composite over a dry season to suppress residual cloud and vegetation flush.
Convert reflectance by dividing the integer DN by 10000 before computing any normalised index.
Generate the interpretation stack: true colour for orientation, 12-11-2 for alteration, B4/B2 and B11/B12 ratios, and edge-filtered B8 for structure.
Bring in topography: load a DEM (Copernicus GLO-30 at 30 m, or national lidar where available), reproject to the scene CRS with gdalwarp -t_srs EPSG:32631, and derive multi-azimuth hillshades.
Interpret and digitise lineaments, contacts and alteration polygons in a GeoPackage, recording confidence and the evidence layer used.
Validate against geological maps and, where possible, field stations.

In Earth Engine, a dry-season median with cloud masking is a few lines:

var s2 = ee.ImageCollection('COPERNICUS/S2_SR_HARMONIZED')
  .filterBounds(aoi)
  .filterDate('2025-06-01','2025-09-30')
  .filter(ee.Filter.lt('CLOUDY_PIXEL_PERCENTAGE', 20))
  .median();
var geology = s2.select(['B12','B11','B2']);
Map.addLayer(geology, {min:0, max:3000}, 'SWIR composite');

Worked example: alteration screen around a structure

Over an arid prospect, a 12-11-2 composite showed a pale, cool-toned halo straddling a NE-trending lineament traced in the edge-filtered B8 layer. The B11/B12 clay ratio confirmed elevated values along the same halo, while the B4/B2 iron ratio peaked on its margins — a pattern consistent with argillic alteration grading outward into an iron-oxide-stained zone. That is enough to high-grade the target for ground follow-up; it is not enough to claim a specific alteration mineral assemblage, which would need ASTER or field spectroscopy.

Common pitfalls and why they happen

Over-claiming mineralogy. Two broad SWIR bands cannot resolve mineral species; the physics simply is not there. Report "hydroxyl alteration likely," not "kaolinite."
Vegetation masquerading as alteration. Dry grass and senescent vegetation have SWIR features that mimic clay tones. Always mask vegetation (NDVI threshold or SCL) first.
Single-azimuth lineament bias. Edges parallel to the sun direction vanish; edges perpendicular are exaggerated. Use multiple hillshade azimuths.
Mixing processing baselines. Sentinel-2 reprocessing introduced an offset (the L2A "BOA add-offset" of -1000 in newer baselines). When comparing old and new scenes, check the baseline metadata or use the harmonized collection.
Treating one cloudy scene as ground truth. Residual haze and thin cirrus distort tones; prefer a clean scene or a dry-season median.

Validation and QA

Confirm the scene CRS and that all derived layers share it; verify cloud masking actually removed clouds by visual inspection at full resolution; cross-check at least one interpreted contact against a published map; and keep a provenance note with tile ID, dates, baseline, indices and band assignments. For lineaments, a second interpreter or a comparison with regional structural data is the cheapest meaningful check.

Bathyl perspective

We position Sentinel-2 as the free, repeatable backbone of a geological remote-sensing programme: it frames the structure, screens for alteration and tells the expensive sensors and the field crew where to concentrate. Its conclusions are always expressed with the confidence the two-band SWIR physics allows.