What is the difference between Landsat and Sentinel-2 for geology?

Sentinel-2 (MSI) gives you 10-20 m optical and red-edge bands with a 5-day revisit, which is excellent for lithological discrimination and vegetation-corrected mapping. Landsat 8/9 (OLI/TIRS) adds an onboard thermal infrared sensor and a 50+ year archive back to 1972, which Sentinel-2 cannot match for change context.

Does Sentinel-2 or Landsat have better SWIR bands for mineral mapping?

Both have two broad SWIR bands at roughly 1.6 and 2.2 micrometres, so neither resolves individual clay or alteration species. Sentinel-2 SWIR is 20 m versus 30 m on Landsat, but for diagnostic mineral absorption you need ASTER's five narrower SWIR bands or hyperspectral data.

Where does ASTER fit for alteration and clay mapping?

ASTER has five SWIR bands (2.145-2.430 um) and five thermal bands, which let you map Al-OH clays, Mg-OH/carbonate, and silica far better than Landsat or Sentinel-2. Its SWIR detector failed in 2008, so usable SWIR data is largely pre-2008 archive.

Which band ratios work on Landsat 8/9 and Sentinel-2?

Classic ratios port across both: ferric iron from red/blue, ferrous iron and gossan from SWIR/NIR, and clay/hydroxyl from SWIR1/SWIR2 (B11/B12 on Sentinel-2, B6/B7 on Landsat 8/9). Always run ratios on surface-reflectance, atmospherically corrected data.

Landsat vs Sentinel-2 for Geological Mapping

Short answer

For geological mapping, Sentinel-2 (MSI) wins on spatial detail and band richness: 10 m visible/NIR, four red-edge bands, two 20 m SWIR bands, and a 5-day revisit from the twin-satellite constellation. Landsat 8/9 (OLI/TIRS) wins on two things Sentinel-2 cannot offer: a continuous land-surface archive reaching back to 1972, and a real thermal infrared sensor (TIRS, ~100 m resampled to 30 m). Neither has the narrow SWIR bands needed to identify specific alteration minerals — for that you reach for ASTER or hyperspectral data. In practice, use Sentinel-2 for lithological discrimination and structure, Landsat for time-series and thermal context, and ASTER SWIR (pre-2008) for clay and alteration mapping.

The sensors, concretely

Sentinel-2 MSI

Two satellites (2A, 2B; 2C now operational) carry the MultiSpectral Instrument with 13 bands and a combined revisit near 5 days at the equator, better at high latitudes. The bands that matter for geology:

B2/B3/B4 (blue/green/red), 10 m — true colour, iron-oxide ratios.
B8 (NIR, 842 nm), 10 m; B8A (narrow NIR, 865 nm), 20 m.
B5/B6/B7 red-edge, 20 m — useful for separating vegetation stress over mineralized ground.
B11 (1610 nm) and B12 (2190 nm) SWIR, 20 m — the two bands you use for hydroxyl/clay and carbonate ratios.

Products are distributed as Level-1C (top-of-atmosphere reflectance) and Level-2A (surface reflectance, atmospherically corrected via Sen2Cor). For any quantitative band ratio, start from L2A or run your own atmospheric correction.

Landsat 8 and 9 (OLI/TIRS)

OLI provides 30 m multispectral and a 15 m panchromatic band; TIRS adds two thermal bands (B10, B11) acquired at ~100 m and delivered resampled to 30 m. Geologically relevant bands:

B2/B3/B4 (visible), B5 (NIR, 865 nm), all 30 m.
B6 (1610 nm) and B7 (2200 nm) SWIR, 30 m — direct analogues of Sentinel-2 B11/B12.
B8 panchromatic, 15 m — pan-sharpen the 30 m stack for visual mapping.
B10/B11 thermal — silica/quartz indices and surface temperature.

The decisive feature is the archive: Landsat 1-5 MSS, then TM (1982), ETM+ (1999), and OLI (2013) give a near-continuous record. For abandoned-mine drainage staining, dune migration, glacier retreat, or any multi-decade change, Landsat has no competitor.

ASTER

Flown on Terra, ASTER carries three VNIR bands (15 m), six SWIR bands (30 m, 1.6-2.43 um), and five TIR bands (90 m). The SWIR set is the point: ASTER bands 5-8 fall across the 2.1-2.4 um region where Al-OH (kaolinite, illite, muscovite, alunite), Mg-OH, and carbonate absorptions live. That lets you build genuine alteration-mineral maps that broad-band Landsat/Sentinel SWIR simply averages over. The catch: the SWIR detector failed in April 2008, so reliable SWIR scenes are pre-2008. VNIR and TIR continue.

Band ratios that port across sensors

The standard geological ratios were designed for Landsat TM and migrate cleanly. Using Landsat 8/9 / Sentinel-2 numbering:

Ferric iron (iron oxides, gossans): B4/B2 (red/blue) — Landsat 4/2, Sentinel-2 B4/B2.
Ferrous iron: B6/B5 (SWIR1/NIR) on Landsat; B11/B8A on Sentinel-2.
Hydroxyl / clay / alteration: B6/B7 (SWIR1/SWIR2) on Landsat; B11/B12 on Sentinel-2.
Iron-oxide composite (Sabins-style): combine B4/B2, B6/B7, B6/B5 into an RGB.

These only mean something on surface reflectance. Run them on L2A (Sentinel-2) or Collection 2 Level-2 surface reflectance (Landsat). Top-of-atmosphere data carries atmospheric scattering that corrupts blue-band-heavy ratios in particular.

Worked example: clay-alteration screen in QGIS + GDAL

Goal: a first-pass alteration screen over an arid target using Sentinel-2 L2A.

Stack the SWIR and NIR bands. Resample B8 (10 m) and B11/B12 (20 m) to a common 20 m grid:
```
gdalwarp -tr 20 20 -r bilinear -t_srs EPSG:32643 B08.jp2 B08_20m.tif
```
(Use the correct UTM EPSG for your tile — e.g. EPSG:32643 for UTM 43N.)

Build the clay ratio B11/B12 with gdal_calc.py:

gdal_calc.py -A B11_20m.tif -B B12_20m.tif \
  --outfile=clay_ratio.tif --calc="A.astype(float)/(B+1e-6)" --type=Float32

Build the ferric ratio B4/B2 the same way, and an RGB composite (ferric, clay, ferrous) in the QGIS layer properties.
Mask vegetation and water. Compute NDVI from B8/B4 and mask cells above ~0.3; mask water with NDWI. Vegetation produces false hydroxyl highs because leaf chemistry also absorbs in the SWIR.
Stretch per-region, not globally. Apply a 2-98% cumulative-cut stretch over your AOI so subtle anomalies are not crushed by an outcrop or a salt pan.

For Landsat, swap B11/B12 for B6/B7 and B2/B4 stay the same; everything else is identical at 30 m.

If you have cloud access, Google Earth Engine does this without downloads: filter COPERNICUS/S2_SR_HARMONIZED by date and CLOUDY_PIXEL_PERCENTAGE, apply the scene-classification mask, build a median composite, then compute ratios server-side.

Resolution, revisit, and cloud trade-offs

Spatial: Sentinel-2 at 10-20 m resolves dykes, narrow units, and structural lineaments that 30 m Landsat blurs. Pan-sharpened Landsat at 15 m partly closes the gap for visual interpretation but not for analysis.
Temporal: Sentinel-2's ~5-day revisit means more cloud-free chances per season — important in tropical or monsoonal terrain. Landsat's 16-day per-satellite (8-day combined 8+9) revisit is sparser but the archive depth compensates for historical work.
Spectral: For lithology, Sentinel-2's red-edge bands help separate vegetation from soil/rock. For mineralogy, the two broad SWIR bands on both sensors only flag "there is some hydroxyl/clay here," not which mineral.
Thermal: Only Landsat (TIRS) and ASTER (TIR) carry it. Silica and quartz-rich units are best discriminated thermally; Sentinel-2 has no thermal band at all.

Common pitfalls and why they happen

Reading false-colour as lithology. A ratio composite highlights spectral contrast, which correlates with but is not identical to rock type. Soil, lichen, desert varnish, and shadow all shift ratios. Validate against mapped geology and, ideally, field spectra.
Mixing TOA and surface reflectance. Comparing a TOA scene to an L2A scene, or running ratios on TOA, injects atmospheric error that mimics iron and clay signals. Keep the processing level consistent and documented.
Expecting mineral ID from broad SWIR. B11/B12 (or B6/B7) span hundreds of nanometres each; kaolinite, illite, and alunite all fall inside the same band and cannot be separated. If the deliverable claims a specific mineral, the data must justify it — that means ASTER SWIR or hyperspectral, not Sentinel-2.
Comparing scenes across season/sun-angle. Different acquisition dates change shadow length, soil moisture, and vegetation, which all move ratios. For change detection, control for phenology and illumination.

QA and validation

Before a remote-sensing layer goes into a report:

Confirm the processing level (L2A / Collection 2 L2) and atmospheric-correction method are recorded in the metadata.
Verify the UTM zone / EPSG is correct for the AOI and that all bands share one grid after resampling.
Cross-check anomalies against at least one independent layer — published geology, geophysics, or known occurrences.
Inspect anomalies at full resolution; many "targets" resolve into roads, tailings, dry riverbeds, or cloud shadow.
State the limitation explicitly: a Sentinel-2 hydroxyl high is an indication, not an assay.

Bathyl perspective

We treat optical satellite data as a vision-extension layer that narrows where to look, not as a substitute for mapping or sampling. The right sensor follows the question: Sentinel-2 for high-resolution lithological and structural screening, Landsat for thermal context and multi-decade change, ASTER for the alteration mineralogy the broad-band sensors cannot resolve. Every layer ships with its processing level, date, and a clear note on what the spectral signal can and cannot prove.

Landsat vs Sentinel-2 for Geology