Sentinel-2's 13 Bands, Explained by Someone Who Actually Uses Them

Someone asked me last week what Sentinel-2 can "really" see. Not the brochure answer. The real one.

So here it is, from someone who's run thousands of scenes across Chagai, Waziristan, and my own concessions in Gilgit-Baltistan. Sentinel-2 has 13 spectral bands. Most people treat them like a black box. They shouldn't.

Honestly, if you understand what even 5 of these bands do, you'll stop getting scammed by consultants selling you "satellite surveys" that are basically Google Earth screenshots with arrows drawn on them.

The Bands Nobody Talks About Properly

Let me walk through this the way I'd explain it to an investor sitting across from me.

Band 1 (443 nm, coastal aerosol). This one's 60m resolution and most analysts skip it. I used to skip it too. Then I realized it's genuinely useful for atmospheric correction over high-altitude sites — and when you're working at 3,800m in Skardu, atmospheric noise wrecks your iron oxide ratios if you don't handle it. So Band 1 earns its keep indirectly.

Bands 2, 3, 4 (blue, green, red — 490, 560, 665 nm). The visible ones. 10m resolution. These give you the true-colour base map, yes, but more importantly they let you calculate the classic iron oxide ratio (Band 4 / Band 2). Rust-coloured gossans that sit above sulphide deposits? They scream in this ratio. In Reko Diq-adjacent terrain, we've flagged oxidation halos using just these three bands before ever touching a shortwave infrared band.

Bands 5, 6, 7 (red edge — 705, 740, 783 nm). These were designed for vegetation. Which sounds useless for mining. But here's the thing — vegetation stress is a proxy for soil geochemistry. Plants growing over copper-rich or serpentinite soils look different from plants 400 meters away on normal substrate. The red edge catches that difference before your eye can. For chromite hunting in Muslimbagh, this is underused and powerful.

Band 8 (842 nm, NIR, 10m) and Band 8A (865 nm, 20m). NDVI lives here. For mining, I use Band 8 inverted — I want to find the bare rock, not the trees. In the Deosai margins where vegetation is sparse anyway, Band 8 helps me separate weathered outcrop from scree.

Band 9 (945 nm, water vapour). Atmospheric correction again. Not glamorous. Necessary.

Band 10 (1375 nm, cirrus). Cloud detection. Skip it for surface analysis.

Where the Real Money Lives: Bands 11 and 12

If you remember nothing else from this post, remember these two.

Band 11 (1610 nm, SWIR-1) and Band 12 (2190 nm, SWIR-2). Both at 20m resolution. These are the shortwave infrared bands and they're the reason Sentinel-2 became serious competition for ASTER for a lot of exploration work.

Why do they matter? Because clay minerals, carbonates, and hydroxyl-bearing minerals — the alteration minerals that sit around ore bodies — absorb light specifically in the 2100-2300 nm window. Band 12 catches the edge of that absorption. The ratio Band 11 / Band 12 is how we map hydrothermal clay alteration across large areas fast.

Porphyry copper systems in Chagai? They have a predictable alteration zonation — potassic core, phyllic (sericite-rich) shell, argillic cap. Sentinel-2 SWIR picks up the phyllic and argillic zones clearly. We've used this on our own concessions to narrow a 40 km² license down to three 800m × 800m drill targets.

For gold, which rarely shows up directly in any satellite band (gold particles are too small and too sparse to produce a spectral signature), the play is always indirect — you're hunting the alteration halo, not the gold itself. Band 11 and 12 ratios are your best friend there.

I got this wrong at first, by the way. Early on at geomines I thought ASTER's SWIR bands (it has 6 of them vs Sentinel-2's 2) would always win. But ASTER's 15m SWIR is noisy now — the instrument has been degrading since 2008 — and Sentinel-2 gives you a new scene every 5 days. Five days. ASTER might give you a usable scene twice a year over the same spot. Revisit frequency beats spectral resolution in most real projects.

What Sentinel-2 Can't Do

Be honest with yourself about this part.

It can't see through cloud. Monsoon season in KP is basically a blackout — that's where SAR earns its salary. It can't see under vegetation canopy. It can't detect lithium directly (lithium brines show up via evaporite mineralogy and surface salts, not lithium itself). It can't tell you grade. And it can't replace a geologist standing on the outcrop with a hand lens.

What it can do is take a 10,000 km² license area and tell you, with reasonable confidence, which 2% of that ground deserves a field team. That's it. That's the entire value proposition. And for a country like Pakistan — where exploration budgets are small and license areas are huge — that 2% filter is worth more than every drone survey combined.

Look, 13 bands sounds like a lot until you realize each one answers exactly one question. Iron oxides. Clay alteration. Vegetation stress. Water content. Atmospheric noise. Stack them right and you get a map nobody else on your block has.

Stack them wrong and you get a pretty picture that convinces an investor to fund a dry hole. I've seen both happen in the last 18 months. Which one do you want to be responsible for?