A vegetation index compares how much red light a surface absorbs against how much near infrared light it reflects. Healthy leaves absorb red strongly and reflect near infrared strongly, so the gap between the two is wide. The index turns that gap into a single number describing how much living leaf is present.
Why leaves look the way they do to a satellite
Chlorophyll absorbs light in the red part of the spectrum and uses it to drive photosynthesis. The internal structure of a healthy leaf, the spongy layer of cells inside it, scatters near infrared light strongly and reflects most of it away.
So a healthy leaf has a very particular signature: dark in red, bright in near infrared. Bare soil reflects red and near infrared much more evenly. Water absorbs near infrared almost completely.
A vegetation index is arithmetic on that signature. The best known, NDVI, takes the difference between the near infrared and red reflectance and divides it by their sum, which produces a value between minus one and one. Dense healthy vegetation sits high. Bare soil sits low. Water goes negative.
Dividing by the sum is what makes it useful across dates. It normalises out much of the variation from sun angle and overall brightness, so an image from March and an image from October can be compared.
What it is good at
It is a screening instrument, and a very efficient one.
It covers every hectare on every pass rather than a sample, which no ground team can match. It has a long archive, so the condition of a parcel before a project started can be reconstructed rather than asserted. And it is cheap, because much of the imagery is free.
That combination makes it the right tool for three jobs: establishing what a site looked like at the outset, spotting which parts of a block are changing differently from the rest, and directing where a ground team should walk.
What it cannot tell you
It measures leaf, not species, not structure and not health below the canopy.
A young plantation and a flush of weeds after rain can return similar values. A dense canopy hides whatever is happening underneath it. Once vegetation is very dense the index saturates, which means further growth barely moves the number, so it becomes a poor guide to biomass in a mature stand.
It is also sensitive to things that have nothing to do with the plants. Cloud and haze, soil colour showing through sparse cover, and the angle of the sun all shift the value.
None of this makes it unreliable. It makes it a partner to ground measurement rather than a replacement for it. The index tells you where to look. A soil test and a survival count tell you what is there.
Reading it over time
A single value is nearly meaningless on its own. What carries information is the shape of the curve.
Compare the same calendar window each year, so seasonal change is held constant. Compare the block against nearby land that was not treated, so a good monsoon does not get recorded as a project result. And watch the dry season value in particular, because anything that holds green through the driest weeks is rooted deeply enough to matter.
That last one is the most useful single reading on restored land. Wet season green is easy. Dry season green is the evidence.
Which index, and why it matters less than it seems
NDVI is the common one. Others exist for specific weaknesses: some adjust for the soil showing through sparse cover, some resist atmospheric haze better, some avoid saturating in dense canopy.
Choosing between them matters far less than choosing a fixed method and holding it. An index applied consistently, on the same dates, over the same boundary, with the working published, produces a record somebody can check. Switching index between reports produces a number nobody can compare to anything.
This is why the method is written down before the measurements start rather than after, which is the same discipline described on the measurement page.
Where this fits
Vegetation index work is one layer of the AI and sensor layer that runs inside every project here. It sits alongside ground counts and soil sampling rather than above them.
See also remote sensing and green cover in the glossary.
Related questions
- Does a vegetation index measure how many trees there are?
- No. It measures how much living leaf surface is present and how vigorous it is. A dense grass cover and a young plantation can return similar values. Counting trees needs either higher resolution imagery where individual crowns are visible, or a ground count.
- Can a vegetation index tell healthy trees from weeds?
- Not on its own. Both are green and both photosynthesise. What separates them is behaviour over time: weeds flush after rain and die back quickly, while established trees hold their value through a longer part of the year. Reading the same site across several seasons distinguishes them far better than any single image.
- How often should it be measured?
- On a fixed schedule, comparing the same calendar window each year. Vegetation changes with the monsoon, so an image from just after rain and an image from late in the dry season describe two different places. Comparing like with like is what makes a change real rather than seasonal.
- What resolution is needed?
- It depends on the question. Free public imagery at tens of metres per pixel is adequate for tracking whether a block as a whole is greening over years. Assessing individual trees, gaps in planting or the boundary of a small parcel needs metre scale imagery, which is usually commercial.