Satellite data¶
Which sensor you have decides which methods you can run, and how specific an answer they can give. SFF needs hyperspectral data. Everything else runs on whatever you ordered, but a broad-band sensor answers a narrower question: it will tell you the ground looks clay-like without telling you which clay, and there are mineral families it cannot reach at all.
Sentinel-2¶
Multispectral, 13 bands at three resolutions: four at 10 m, six at 20 m, three at 60 m. It arrives with every area.
The 60 m bands exist for atmospheric correction and cirrus detection rather than for surface work. The cirrus band is dropped from the surface-reflectance product; the coastal-aerosol and water-vapour bands are still present and currently enter the analysis, and removing them is planned. Everything you will actually use is 10 or 20 m, with both SWIR bands at 20 m.
What it can do. Sentinel-2 carries two SWIR bands, B11 near 1610 nm and B12 near 2190 nm, and B12 is 180 nm wide. That is enough to flag Al-OH and clay-like ground from the B11:B12 ratio, and enough to map iron oxides in the visible and near infrared.
What it cannot do. It cannot name a mineral, because the features that separate kaolinite from illite are far narrower than the bands. More importantly, it does not reach carbonate. B12's response ends near 2290 nm, while the calcite and dolomite features sit at roughly 2320 to 2340 nm, and there is no band beyond B12 to form a ratio against. An Al-OH absorption at 2200 nm and a carbonate absorption at 2340 nm both just darken the same wide band. For carbonate or Mg-OH work you need ASTER or hyperspectral data.
The advantage is the pixel and the revisit. At 10 to 20 m you can see an outcrop that a 60 m pixel would average away, and the archive refreshes every few days, so you can pick a dry, snow-free quarter.
EMIT¶
Hyperspectral: 285 contiguous bands from about 380 to 2500 nm, at roughly 60 m per pixel. Ordered on request.
The continuous spectrum is what makes named-mineral work possible, because absorption features can be fitted individually instead of inferred from band ratios. SFF needs data of this kind, and where you have it SFF is the most reliable mineral match in the platform.
The latitude limit. EMIT flies on the International Space Station, so it only ever observes between about 52°N and 52°S. Above or below that band there will never be EMIT data, however long you wait, and SFF is unavailable on that ground. That excludes most of Canada, Fennoscandia and northern Russia. Inside the band, coverage still depends on what the instrument actually acquired, since its observation plan was built around arid dust-source regions. The Status tab tells you when there is nothing usable for your area.
The pixel. One EMIT pixel covers about 0.36 hectares and reports a single spectrum for all of it: rock, soil, dry grass and shadow, mixed together. A real absorption feature from a small exposure survives that mixing as a shallower version of itself, which is why the curve check reads where a dip sits rather than how deep it goes.
Other hyperspectral sources, EnMAP and PRISMA, also support SFF where they are available for your ground.
ASTER¶
Multispectral: three VNIR bands at 15 m, six SWIR bands at 30 m, five TIR bands at 90 m.
The SWIR bands are the ones that matter for mineral work, and they sit at 30 m: coarser than Sentinel-2's 20 m, finer than EMIT's 60 m. Two of them fall between about 2.295 and 2.365 µm, which is why ASTER can discriminate carbonate and Mg-OH minerals that Sentinel-2 cannot reach.
The archive constraint. ASTER's SWIR bands have been saturated since April 2008, after the instrument's cooling system degraded, and quality was already falling through 2007. Treat the usable SWIR archive as pre-2007. ASTER's VNIR and TIR bands still acquire today, but the mineral diagnostics are in the SWIR.
So anything you read from ASTER SWIR describes the surface as it was in 2007 or earlier. On stable ground that is often fine. Over anything mined, farmed, built on or eroded since, it is a historical record rather than a current observation, and you should date it explicitly when you write your target up.
Bare Earth Composite (BEC)¶
A composite built from a time-series rather than a single acquisition. Ordered on request.
BEC looks across many dates and keeps, for each pixel, the observation where that ground was least covered. In vegetated terrain this recovers exposures that any single scene would hide behind canopy or crop.
Two consequences to plan around:
- The scene has no single date. Pixels come from different acquisitions, so BEC cannot carry an acquisition date the way a normal scene does. Record it as a composite when you write down what produced a target.
- It ships already masked. The masking is part of how the composite is built, so you cannot apply your own mask on top of it. Build masks for your other providers and use BEC as delivered.
Order it to complement Sentinel-2 in vegetated ground, not to replace it.