Knowledge base¶
This tab is the reference half of the documentation. How to work in TerraEye tells you what to click and in what order. The knowledge base tells you what each layer actually computes, which sensor it needs, and where it will mislead you.
It splits three ways, and the split is the same one the platform uses: what you are given, what is done to it, and what you interpret it with.
| Section | Answers |
|---|---|
| Data | What arrives with your area, what you have to order, and what each source can physically resolve. |
| Methods | What each algorithm computes, which way its values run, and how it produces a false positive. |
| Laboratories | The two tools you interpret with: Analogy Lab and Data Masking. |
Data¶
What arrives with your area and what you have to ask for. Pixel sizes, wavelength coverage, acquisition dates, grid resolutions, and the upload formats we accept.
Methods¶
The algorithms that turn that data into layers. Each page states what the method measures, which way its values run, and the conditions under which it returns a false positive.
Laboratories¶
The two tools you interpret with: Analogy Lab for spectral curves and similarity maps, Data Masking for removing surfaces that drown out mineral signal.
Reading a layer without getting burned¶
Four things hold across every method here, and they cause most of the wrong conclusions we see.
A satellite reads the exposed surface layer only. What you map is surface mineralogy. Part of it may be hydrothermal alteration, spread wider than the deposit that caused it; much of it is weathering product, regolith, or unaltered rock carrying the same clays and iron oxides. Barren alteration also vastly outnumbers mineralized alteration, so a mapped halo is a search area and a weak prior, not a find.
Cover decides whether any of it means anything. Residual regolith weathered in place carries a diluted, weathering-modified version of the bedrock signature beneath it. Transported cover — till, alluvium, colluvium, wind-blown sand — describes itself and tells you nothing about what lies under it. Canopy and standing water are opaque either way. Knowing which regolith regime your ground is in is a precondition for reading any spectral result, and it is the same distinction that decides whether a soil assay means anything. Magnetics can see through cover, but only where you have a survey fine enough to resolve your target.
Pixel size sets the smallest thing you can trust. One EMIT pixel averages 0.36 hectares of mixed rock, soil and shadow. A mineral present in a tenth of that pixel may not survive the mixing.
Same-image agreement is one observation, not several. SFF, SAM and GSI run on the same pixel of the same scene. When they agree, three algorithms agree about one measurement. That is still worth having, because the three fail differently and agreement rules out method-specific artifacts, but it adds no independent observation of the ground. Different providers and different years are less correlated and worth more. The only genuinely independent test is a physical one: a sample assayed, a field spectrometer reading, or a hole.
Every method returns false positives. The curve check in Analogy Lab is the step that separates a result from a lead, which is why Phase 7 exists.