NDVI is the normalised difference vegetation index, the most widely used remote sensing index in existence. It is computed from two bands:
NDVI = (NIR - Red) / (NIR + Red)
where NIR is reflectance in the near infrared, typically around 800 to 900 nanometres, and Red is reflectance in the red, typically around 650 to 680 nanometres. Values range from -1 to 1.
Why those two bands
Healthy green vegetation has a distinctive spectral shape. Chlorophyll absorbs strongly in the red, so red reflectance is low. Internal leaf structure scatters near infrared light very efficiently, so NIR reflectance is high, often above 0.4. The gap between the two is large and it is specific to living plant tissue. Almost nothing else on a landscape produces that combination.
Normalising by the sum rather than using the raw difference makes the index far less sensitive to illumination differences, slope, and overall brightness. A shaded slope and a sunlit slope with the same vegetation return similar NDVI values even though their absolute reflectances differ.
Typical values are informative in a rough way. Dense healthy canopy sits high, commonly above 0.6. Sparse or stressed vegetation sits in the middle. Bare soil sits low, usually below 0.2. Water and snow can go negative.
What NDVI does not tell you
It does not measure biomass, yield, chlorophyll content, or plant health directly. It correlates with all of them under some conditions and none of them reliably across conditions.
It saturates. Above roughly a leaf area index of three, adding more canopy barely changes the value, so NDVI stops discriminating exactly where dense crops and forests live.
It is sensitive to soil background in sparse canopies, which is why soil adjusted variants exist.
It does not identify species, and it cannot separate a crop from a weed with similar canopy structure.
It tells you nothing about cause. A drop in NDVI is consistent with drought, disease, harvest, flooding, or a cloud shadow that survived masking.
Treated as a screening layer that flags where to look, NDVI is excellent and cheap. Treated as a measurement of a physical quantity, it will mislead. Where the question is what changed and why, narrow band indices targeting specific pigments and water absorption features, or full spectrum methods, carry considerably more information.