Band Ratio Mineral Exploration: ASTER, Landsat, Sentinel-2

A band ratio divides the brightness of one satellite band by another, pixel by pixel, so that a mineral’s absorption feature stands out and sunlight and shadow cancel. The three ratios used most often in exploration are red divided by blue for iron oxides, the first shortwave-infrared band divided by the second for clay and other hydroxyl minerals, and a near-infrared or shortwave pair for ferrous iron. On Landsat 8 and 9 the first two are 4/2 and 6/7; on Sentinel-2 they are B4/B2 and B11/B12; on ASTER they are 2/1 and a set of ratios built from bands 4 to 9.

This page gives the formula for each sensor, what the result maps, the colour composites that geologists combine them into, a six-step workflow, and the limits that decide whether a bright pixel is worth a field visit. For the wider picture of satellite work on a licence, see our guides to satellite mineral detection, remote sensing for mineral exploration and AI in mineral exploration.

Band ratio formulas for iron oxide, clay and ferrous iron

The table lists the ratios reported in peer-reviewed mapping studies. Each one uses the same physics: iron oxides absorb blue light and reflect red, while clay, mica and other hydroxyl-bearing minerals absorb near 2.2 micrometres in the shortwave infrared (SWIR). The sensor bands change, the idea does not.

Sources for the table: the Landsat 8 and ASTER ratios come from a lithological mapping study in the High Atlas of Morocco (ISPRS Archives, 2019); the kaolinite and alunite ratios come from an ASTER test at Biga, Turkey (ISPRS Congress, 2004); the Sentinel-2 ratios come from a volcanogenic massive sulfide study in Eritrea (PLOS One). Authors write some ASTER ratios in different ways, so check the band numbers against your own data before copying a formula into software.

ASTER, Landsat 8 and 9, and Sentinel-2 bands compared

ASTER, the Advanced Spaceborne Thermal Emission and Reflection Radiometer on NASA’s Terra satellite, has 14 bands: three visible and near-infrared bands at 15 m, six shortwave-infrared bands at 30 m and five thermal bands at 90 m, over a 60 km swath (NASA/JPL ASTER characteristics). Its six SWIR bands sit at 1.600-1.700, 2.145-2.185, 2.185-2.225, 2.235-2.285, 2.295-2.365 and 2.360-2.430 micrometres, which is why geologists value it: the narrow slices straddle the absorption features of kaolinite, alunite, sericite and chlorite.

Chart

Shortwave-infrared bands on each sensor

Number of bands between 1.5 and 2.5 micrometres

Landsat 8 and 9 carry the Operational Land Imager, with nine bands (eight at 30 m, one panchromatic band at 15 m) including two SWIR bands, 1.57-1.65 and 2.11-2.29 micrometres (NASA Landsat spectral bands). Landsat 9 launched on 27 September 2021 and carries enhanced versions of the Landsat 8 instruments, so the same ratios apply to both (NASA Landsat 9). Sentinel-2 samples 13 bands: four at 10 m, six at 20 m and three at 60 m, with the two SWIR bands near 1610 and 2190 nanometres at 20 m (ESA SentiWiki, Sentinel-2 mission).

Two things follow. ASTER has no blue band, so its iron-oxide ratio uses red over green (2/1) instead of red over blue. And only ASTER can split the 2.2 micrometre feature into narrow parts; the two-SWIR-band sensors give one clay signal, not a clay type.

Iron oxide ratios: what they map and what they miss

Iron oxide and hydroxide minerals such as hematite, goethite, limonite and jarosite absorb in the visible and near-infrared range, 0.4 to 1.1 micrometres, and give weathered rock a yellow or red colour. Where sulfides oxidise at surface the stained cap is called gossan, and gossans over porphyry copper and massive sulfide bodies are a classic first target (review of ASTER, ALI and Hyperion data for ore exploration). The ratio works because red light is strongly reflected and blue light strongly absorbed, so red over blue is high on iron-stained ground.

Four things reduce its value in practice:

• Red soil and laterite look the same. In the Asmara belt of Eritrea, hematitic laterite is spectrally close to true gossan in broadband data. The authors separated them by combining the iron signal (4/2) with the clay signal (11/12): gossans scored high on both, laterites mostly on the clay signal (PLOS One).
• Iron is not ore. Many barren rocks carry iron stain, and many ore bodies have none at surface.
• Cover hides it. Soil, transported gravel and vegetation block the signal; only exposed, weathered ground counts.
• Shadow and bright surfaces change absolute brightness. Ratios cancel most of this, but steep slopes and salt crusts still leave artefacts.

▶ Satellites cannot see gold; they map rusty iron stains, clay minerals and cracks left by hot water. · 4:25

Clay and hydroxyl ratios from the shortwave infrared

The shortwave infrared is the best region for hydrothermal alteration, because Al-OH, Mg-OH, Si-OH and carbonate groups absorb there. Sericite and muscovite show a main Al-OH feature near 2.2 micrometres and a weaker one near 2.35; kaolinite and alunite, typical of advanced argillic alteration, show Al-OH features at 2.165 and 2.2; chlorite, epidote and calcite in propylitic zones absorb between 2.31 and 2.33 (review of ASTER, ALI and Hyperion data).

Chart

Pixel size by sensor

Ground sampling distance of visible and shortwave-infrared bands, metres

A clay ratio puts a band on the shoulder of that dip (reflectance high) over a band inside it (reflectance low). On Landsat 8 and 9, band 6 is on the shoulder and band 7 covers the dip, so 6/7 is bright over kaolinite, montmorillonite and other clays. On Sentinel-2, B11 is the shoulder and B12 the dip, so B11/B12 does the same job at 20 m. The signal marks where hydroxyl minerals are abundant; it does not say which mineral, and it does not say whether the alteration is related to ore. A clay-rich shale and a sericitic alteration halo can give the same ratio value.

For a worked treatment of alteration zones from porphyry copper to epithermal gold, see our guide to alteration mineral satellite mapping.

ASTER band combinations that separate clay types

This is where ASTER earns its place. Because bands 5, 6, 7 and 8 are narrow, ratios between them respond to different minerals. In the Biga district of Turkey, researchers applied 4/5 to find alunite and 7/6 to find kaolinite; across 19 hydrothermal sites the ratio method identified kaolinite at 11 and alunite at 8, and the authors concluded that ratios suit kaolinite better than alunite, while a spectral-index method did better on alunite (ISPRS Congress, 2004). In the Moroccan study, (5+7)/6 was used for smectite and illite, a relative band depth of (7+9)/8 for carbonate and chlorite, and 2/1 for ferric iron.

The tests are small and local: Biga is one district. A ratio that separates alunite from kaolinite there may need new thresholds in a different climate or rock type, so treat the table as a starting list, not a recipe.

Ferrous iron, carbonate and colour composites

Single ratios are hard to read in grey. Geologists load three ratios into the red, green and blue channels of one image so that each mineral group takes a colour. Studies on each sensor use set recipes:

• Landsat 8, Sabins ratios: 4/2, 6/7 and band 10 as RGB, or 4/2, 6/5 and 6/7 as RGB; the Moroccan study also tested 6/7, 4/2, 5/4 (ISPRS Archives, 2019).
• Sentinel-2, Abrams ratios: 4/2, 3/1 and 11/8 as RGB, which confirmed iron-oxide anomalies over known gossans in Eritrea.
• Sentinel-2, Chica-Olma ratios: 11/12, 11/8 and 4/2 as RGB, which show clay alteration, ferrous minerals and iron oxides in separate colours (PLOS One).

The Eritrean team also classified rock units from the same Sentinel-2 scenes and reached overall accuracies of 73.34 percent and 73.04 percent at its two test sites, which is a useful reminder: a good colour image is a screening aid, and mapped units still need a geologist’s check.

How to build a band ratio map in six steps

The workflow is the same on any sensor. Only the band numbers change.

Infographic

Band ratio map in six steps

1. 1Pick a sceneDry season, little cloud, recent date.
2. 2Use reflectanceSurface reflectance, not raw brightness.
3. 3MaskRemove vegetation, water and cloud with NDVI and NDWI.
4. 4Divide bandsIron: red over blue. Clay: SWIR 1 over SWIR 2.
5. 5CombineLoad three ratios into one RGB image.
6. 6Check on the groundCompare with the geology map and samples.

1. Pick a dry-season scene with little cloud. Vegetation and moisture reduce rock signal. Landsat 8 and 9 repeat every 16 days each and Sentinel-2 every five days with the full constellation, so most licences have several candidates per season; the harmonized Landsat and Sentinel-2 product merges them to 1.4 days for five satellites (NASA Earthdata, HLS).
2. Use surface reflectance, not raw brightness. Sentinel-2 Level-2A and the harmonized Landsat and Sentinel-2 product are supplied as surface reflectance, so atmospheric effects are already corrected (Earth Engine catalog, Sentinel-2 Level-2A).
3. Mask vegetation, water and cloud. The Eritrea study applied NDVI and NDWI thresholds before computing ratios so that only geologically relevant pixels remained.
4. Compute each ratio. Any raster calculator can divide one band by another; resample Sentinel-2 so both bands in a ratio share a pixel size.
5. Stretch and combine. Clip the extreme tails, then build a three-ratio RGB composite.
6. Check against ground truth. Overlay the government geology map, known workings and any soil or rock samples, and drop anomalies that fall on roads, quarries, burnt ground or red alluvium.

Limits: where band ratios mislead

Band ratios are cheap and quick, and they are blunt. The main failure modes are:

• Broad bands. Landsat and Sentinel-2 bands are tens to hundreds of nanometres wide, so mixed pixels and similar minerals blend. Two SWIR bands cannot tell kaolinite from alunite or sericite.
• Cover. Hydroxyl and iron signals come from the top few micrometres of exposed rock or soil. Under forest, snow or a metre of transported cover they say nothing about bedrock.
• False positives. Weathering, salt, burn scars, mine tailings and agriculture produce clay-like or iron-like pixels.
• Alteration is not mineralization. Large areas of altered rock carry no metal. Ratios shortlist ground; samples and drilling prove it.
• No grade or depth. A ratio value is not a grade. Satellite results are exploration targets that sit ahead of a resource estimate under JORC or NI 43-101.

▶ Spectral signatures explained: what hyperspectral and multispectral imaging can and cannot tell you. · 1:49

The ASTER process and the shortwave-infrared gap

The standard geometrically corrected product is Level 1T. It holds calibrated at-sensor radiance, the same quantity as Level 1B, that has been geometrically corrected, rotated to a north-up UTM grid and resampled once from the Level 1A data. The precision terrain correction uses GLS2000 elevation data and ground control points on daytime scenes (NASA LP DAAC, AST_L1T).

The key caution for any mineral map: ASTER’s SWIR detectors have not given usable data since April 2008. NASA’s data centre states that SWIR data acquired from April 2008 onward show anomalous saturation and striping, and the SWIR bands are no longer included in the Level 1T product. The six-band clay and carbonate ratios in this guide therefore need an archive scene from March 2000 to April 2008, and NASA notes striping in some earlier scenes too, so check each scene’s SWIR quality. Visible and thermal bands continue, but a failure of a power-transmitting unit on Terra on 28 November 2024 left too little power for the instrument; visible and near-infrared imaging resumed on 18 January 2025 and thermal imaging on 15 April 2025 (NASA LP DAAC). For fresh clay mapping, Landsat 9, Sentinel-2 or hyperspectral data now do that work, at lower mineral detail.

Choosing a sensor and where to get the data

Match the sensor to the question and the scale of the licence. The chart shows why wide swaths matter for regional screening and narrow ones for detail.

Chart

Swath width of one scene

Kilometres across track

Sentinel-2 data are published through the Copernicus Data Space Ecosystem, and ASTER and Landsat archives through NASA and USGS data centres. Many teams compute ratios in a cloud platform such as Google Earth Engine, or in desktop GIS, because the arithmetic is simple.

Band ratios and rare earth, copper and gold targets

Copper. Porphyry copper systems have a concentric pattern of potassic, phyllic, argillic and propylitic alteration, and iron-oxide caps over sulfides. That pattern is why band ratios found their first use on porphyry districts (review of ASTER, ALI and Hyperion data). For context on where the metal ends up, see the five largest copper producers.

Gold. No satellite sees gold. Ratios map the iron stain, clay and silica of the system around it, mostly in epithermal and orogenic settings with exposed rock. Where forest, snow and lakes hide bedrock, as across boreal Canada, results thin out, which matters to explorers listed in our guide to Canadian gold mining companies.

Rare earths. No broad-band ratio measures a rare earth element directly. At best a ratio flags the host rock or alteration around a deposit, and the map then guides sampling. If you are new to the group, start with what a rare earth metal is, the list of the 17 rare earth metals, the critical minerals definition and neodymium and radioactivity.

Farmonaut’s mining service applies multispectral and hyperspectral analysis to a licence boundary you supply as a KML, KMZ or polygon, and delivers a report in 5 to 20 business days depending on size and mineral. Use it to screen ground before you spend on field work.

Frequently asked questions

What are the ASTER satellite bands?

ASTER has 14 bands: three visible and near-infrared bands at 15 m, six shortwave-infrared bands at 30 m (1.60 to 2.43 micrometres) and five thermal bands at 90 m, over a 60 km swath, as listed by NASA/JPL. The SWIR bands do the mineral work; they have returned unusable data since April 2008.

What is the best band ratio for iron oxide?

Red over blue: 4/2 on Landsat 8 and 9, B4/B2 on Sentinel-2. ASTER has no blue band, so 2/1 (red over green) is used for ferric iron.

What is the best band ratio for clay minerals?

SWIR 1 over SWIR 2: 6/7 on Landsat 8 and 9, B11/B12 on Sentinel-2. ASTER users add (5+7)/6 for smectite and illite, 7/6 for kaolinite and 4/5 for alunite.

Which ASTER band combinations are used for alteration mapping?

Studies use ratios of bands 4 to 7 for clay types, a relative band depth of (7+9)/8 for carbonate and chlorite, and 2/1 for ferric iron. The table in this guide lists each, with the mineral it targets.

What does the ASTER process involve?

Level 1T is calibrated at-sensor radiance, geometrically corrected, rotated to a north-up UTM grid and resampled once from the Level 1A data, using elevation data and ground control points. You then convert radiance to reflectance before computing ratios.

Can I use Landsat 9 for the same ratios as Landsat 8?

Yes. Landsat 9 carries enhanced versions of the Landsat 8 instruments with the same bands, so 4/2, 6/7 and 6/5 apply to both and the two satellites double the number of scenes.

Is Sentinel-2 enough for mineral mapping?

For iron-oxide screening and a general clay signal, yes. It has 10 m visible pixels and a five-day revisit but only two SWIR bands, so it cannot split clay types the way ASTER archive data can.

Do band ratios find gold or copper?

No. They map iron stain and hydroxyl alteration that often surrounds gold and copper systems. The result is a target for sampling, not a resource estimate.

Do band ratios work under vegetation?

Poorly. Mask vegetation first with NDVI, and expect gaps in forest, farmland and wet ground. Dry, sparsely vegetated terrain gives the best results.

How large is the area one scene covers?

One ASTER scene covers 60 by 60 km, a Landsat scene 185 km across and a Sentinel-2 swath is 290 km wide (NASA Landsat 9, ESA). A single scene therefore covers most licences.

Do I need a licence or permission before sampling a satellite anomaly?

Yes. Check the claim and land-access rules with the regulator and the land owner before any fieldwork, and never use mercury for sampling or recovery.