How is satellite data used to find drug crops?

Question

Almost all leaves are a pretty similar shade of green, and from the altitude of a satellite a single pixel must be bigger than a whole bush. Yet satellite imagery has become central to detecting illegal drug crops - even though farmers take extensive steps to hide the crops by interspersing the plants among legal crops and planting small, irregular fields.

The World Drug Report 2010 explains a little, but also makes it sound extremely hard:

Satellite images contain more information than a normal photo, since the satellites have extra sensors to register infrared colours. Precisely these infrared colours help to distinguish vegetation types. [From footnote:] Coca is a bush and the leaves can be harvested throughout the year. Therefore, the crop stage can vary by field but also within a field. At the same time a satellite image is taken, a field can contain different crops in different growing stages, which give a different appearance in the satellite image.

So it has something to do with infrared information, but how does that give such specific information that an individual kind of plant can be distinguished with any confidence at all, even despite efforts to hide it, and confounding factors such as weather, slopes, and a wide range of mixed developmental stages?

Answer 1

If you're not asking specifically about the near IR range photogrammetry but more in a general sense, then one obvious method of detection of covert drug plantations is simply human activity around them and lack of any public records about activities in the area. So even satellite imagery in the visible spectrum could reveal access roads, air strips, piers and traffic in the middle of nowhere. Imagery in the IR range could then help detect such activities also during nights, reveal hot spots where electricity or open fire are used extensively, even that vehicles at rest have recently been on the move. A coffee or peanuts plantation simply won't have same patterns of activity than a coca farm.

In near IR spectrum though, it's possible to distinguish between plant species in a similar way than we can all see the difference in foliage colors between certain tree species at certain seasons of the year, the difference only becomes much more apparent in near IR range because that spectrum extends to where plant species differentiate more than it's all green to me in the visible spectrum. On average, foliage photosynthesizes most intensively in the yellow to orange spectrum, i.e. that's the light frequency they absorb, while they reflect the rest and that's why they appear to us (we see reflected light but not absorbed one) as the exactly contrasting dark green.

But plants at different stages of maturity also shift this absorbed spectrum. Most obvious is the shift to lower frequency (more red and into IR range) absorption during flowering. So if you know what season it is somewhere, and when cultures flower in that season, you can then extrapolate what crops you might be looking at. Couple this with growth rate (by measuring thermal inertia of the area you can estimate vegetation density), and you're one step closer to identifying one plantation's purpose. In other words, plant species' biological markers are more distinct if you can also include observations in the IR / near IR range.

Repeat observations will then give you more clues, up to the point when you decide that you need boots on the ground.