One of the most powerful uses of remote sensing is being able to assess vegetative growth. This article will explain why remote sensing is so powerful when working with vegetation, and explain NDVI and a few other indexes of vegetation growth. To see how to use Radiant.Earth to perform those calculations, look at the example in Creating your first template and analysis.
Light reflects in a broad spectrum, and every object reflects different parts of the spectrum at different intensities. This means that every object has a specific spectral signature. We can see some of this with the naked eye: a blue crayon reflects more blue light than other parts of the visible spectrum, distinguishing it from the red crayon.
But visible colors are only part of the story. Different objects reflect invisible light at different intensities too.
How does this help with vegetation? We think of a forest as mostly green, so we might assume that trees reflect green light the most. In reality, vegetation reflects most highly in the Near Infrared part of the light spectrum. This spectral signature is fairly unique to vegetation, so we can use this knowledge to our advantage in determining the extent of healthy vegetation.
In this graph, we can compare how water, soil, and vegetation each reflect light. Water doesn't reflect high at all. Soil reflectance goes up slowly as the wavelength increases. But vegetation reflects fairly low in the visible spectrum (a little higher in the green, which is why it appears green to us) and very high in the NIR.
One quick way of seeing this difference is comparing a "True Color" image with a "False Color" image. The true color layers the three visible bands (Red, Green, and Blue) to make an image that looks like what we might see with a naked eye: water appears dark blue, vegetation appears green, etc. However, you might notice in this picture that distinguishing between vegetation and water isn't always that easy: they look pretty similar.
This scene of the Florida pan handle, for example, has serious contrast issues in the visible spectrum:
If, however, we display in false color, we display the Near Infrared band as red, the Red as green, and the Green as blue (blue and green are usually similar enough that they mesh well together). You'll notice that just about everything looks the same, except that now vegetation is lit up in red, making the vegetation-covered land much easier to identify.
We can use this to our advantage in getting some finer calculations about vegetation. These calculations take advantage of the fact that plants reflect high in the NIR, but low in the Red (which is why they look green rather than white). This helps to distinguish between plants and objects that are reflecting high across the spectrum (concrete is bright in the Red and NIR).
The most basic function is a Simple Ratio. It just compares the Red and NIR bands, on the assumption that anywhere the NIR is high and the Red is low, there's vegetation. The ratio is calculated fairly straightforward:
NIR / Red
However, the Simple Ratio can only work on one image, where all of the lighting is exactly the same. You can't compare the ratio calculated in one image to the ratio calculated in another image, because one image might be brighter than the other, producing an overall higher ratio.
Normalized Difference Vegetation Index (NDVI) was established to counter this problem. It compares the difference between the NIR and Red bands to the total reflected light from both bands, normalizing the relationship, which is calculated with this formula:
(NIR - Red) / (NIR + Red)
This formula will result in an index between -1 and 1, which means that different images, even from different parts of the world, will be comparable. When performing a classification, a threshold is established: often, anything with an NDVI of 0.5 or higher is considered vegetation.
This calculation can be used to identify healthy vegetation and track change over time, highlighting drought or surplus.
Both NDVI and Simple Ratio can be calculated and displayed using Radiant.Earth. If you want to see how powerful NDVI is, follow this tutorial to calculate it on your own projects.