Skip to content
Anora

Learn · Foundations

The concept of an index

Two bands, one subtraction and one division. It is the smallest piece of arithmetic in remote sensing and it is the step where an image turns into a number you can put in a report.

This follows on from how light works, which is worth reading first if you have not.

Loading bands…
Red · 665 nm what the sensor recorded
 
 
 
665 nm, then 842 nm

Two pictures, taken at the same instant

This is a city with a river through it, planted banks, three parks and a cluster of towers. Below is the same city twice, one wavelength at a time, from the same pass on the same day.

In red light the parks sit at about four percent and the rooftops at thirty one, so the greenery is the darkest thing in the frame. Move to the near infrared and the parks jump to fifty percent while the rooftops barely shift. Nothing on the ground changed between these two pictures. The only difference is which wavelength we kept.

Near infrared − red

Subtract one from the other and the greenery separates itself

Take the two images and subtract them pixel by pixel. Anything that reflects both bands about equally cancels to nothing, which is most of the built fabric. Anything that is dark in red and bright in infrared survives, which is almost exactly the definition of a healthy leaf.

The parks come out at about +0.46 and the rooftops at +0.03. That is already useful, and it is already fragile. A subtraction carries the brightness of the day it was taken, so the same park photographed under thinner light would give a smaller number and look like less park.

The index

Divide by the sum and the number stops depending on the light

Divide the difference by the total of the two bands. That single step is what turns a measurement of this image into a measurement of the ground, because whatever scaled both bands equally now appears on the top and the bottom of the fraction and cancels out.

The result is the Normalised Difference Vegetation Index, and every index in the next piece is this same shape with different bands in it. The parks read +0.86, the rooftops +0.05 and the river −0.50.

The same ground, the sun coming down

Drop the sun and the brightness collapses. The ratio does not.

Here the wavelength is held still and the sun is walked down the sky, from sixty two degrees above the horizon to sixteen. Watch the parks. As the towers throw shadow across them their brightness falls to under sixty percent of where it started, and a threshold set on brightness alone would begin quietly reclassifying green as something else.

The second figure is the one to watch. It is the ratio of infrared to red over that same green, measured on these same darkening frames and quoted against where it began. Brightness loses more than forty percent. The ratio holds within three. That ratio is what an index is built from, and it is why an index survives a change of light that raw brightness cannot.

Reading the number

Now the number means something you can act on

An index is only worth having if its values map onto the world. Open water sits below zero. Asphalt, concrete and bare ground cluster near zero. Grass runs around 0.3 to 0.5, and closed healthy canopy sits above 0.8.

That is what turns a picture into a register: green cover per ward, measured the same way each time, comparable between one season and the next. Not because the image looks greener, but because the number is defined and the light has been divided out of it.

Why the division is the part that matters

Almost anyone can be talked through the subtraction. The division is the step that gets skipped, and it is the one doing the real work. Sunlight varies with the season, the hour, the haze and the slope of the ground. All of it scales both bands together, so all of it divides out.

That is why a value from one date can be compared with a value from another, and why a shaded courtyard and a sunlit park can be held to the same standard. Without it you have a picture that happens to be numeric. With it you have a measurement.

Where we use this

Municipal green cover registers are the most direct application: canopy and green space by ward, tracked between seasons and between years, with a number attached rather than an impression. At ten metre resolution that is park and block level cover, which is the right scale for planning decisions and for holding a target to account.

The same arithmetic underpins agricultural stress mapping, deforestation screening and water extent. What changes is which two bands go into it, which is the subject of the next piece.

Reflectance values here are representative figures for each surface type, not measurements of a specific place.

Put this to work
Bring us a question about your ground
If you need green cover, crop condition or water extent measured rather than estimated, this is the arithmetic underneath it and field validation is where we finish.
Get in touch →