Dividing by NIR + RED creates a normalized value whereby NDVI will always fall between -1.0 and +1.0. This narrow range of values lends itself to a rapid assessment of the ground cover present in your data. Clouds, water and snow tend to have very small positive values or even negative values; bare soils tend to have small positive values (perhaps 0.1 to 0.2); while health vegetation tends to have values of 0.3 or greater.

You may note that the word “tend” has been used in all instances above. This was done on purpose as this should be taken only as a broad-brush attempt at categorization. NDVI calculations will vary for many reasons that are not directly related to the health of the plant itself. Several confounding factors include light atmospheric haze which obscures surface reflection; sensor (e.g. QuickBird, IKONOS) optics which are never the same so each measures slightly different digital numbers for the NIR and RED bands over the exact same area; and the geometric relationship of the plant to the sensor as well as the shape of the plant itself as this alters surface reflectance angles.

While certainly not an exact science, NDVI is another tool imagery users can put in their back pocket when tackling a project with a focus on vegetation. NDVI is an easy way to separate green, healthy plants and trees from surfaces that are human-made, bare or are covered in water/snow. NDVI can also identify pockets of less healthy crops in fields which can be an early indication of poorly irrigated plants, unbalanced soil nutrients or a pest infestation. City officials have used NDVI to identify watering restrictions violators in the midst of severe droughts by locating pockets of abnormally healthy vegetation. As is true with most technological applications, the usefulness of NDVI is only limited by the imagination of the user…

NDVI calculation for an urban landscape where green patches depict healthy vegetation while yellow, orange and red areas are bare soils or human-made structures.