NDVI & Vegetation Indices

A green leaf does two revealing things with light. Its chlorophyll absorbs red light to drive photosynthesis, so vegetation is surprisingly dark in the red band. But the internal cell structure of a healthy leaf strongly reflects near-infrared (NIR) light, which we cannot see but a satellite sensor records. So healthy vegetation is dark in red and bright in NIR - a big gap between the two bands.

Stressed, dying or absent vegetation closes that gap: less chlorophyll means less red absorption, and collapsing cell structure means less NIR reflectance. Bare soil, water and concrete show little or none of the vegetation pattern at all. This single, reliable contrast between red and NIR is the physical fact every vegetation index exploits - no botany degree required, just two bands.

NDVI puts the red-NIR gap into a formula: NDVI = (NIR - Red) / (NIR + Red). The top measures the gap; dividing by the sum normalises it, which is the clever part. Because it is a ratio, NDVI largely cancels out differences in overall brightness - a slope in shade and the same vegetation in full sun give similar NDVI values, even though their raw band numbers differ. That is what makes NDVI comparable across a scene and, with care, across dates.

The result always falls between -1 and +1. Open water sits below zero; bare soil, rock and built-up surfaces cluster near zero (roughly 0 to 0.2); grass and crops run mid-range; dense, healthy forest or a well-watered park pushes toward 0.8. You are not reading absolute biomass - NDVI saturates over very dense canopy and is sensitive to soil background when cover is sparse - but as a relative greenness measure it is fast, free and remarkably robust.

You need one red band and one NIR band, and every major optical mission has both. On Sentinel-2 (free) the red band and the NIR band are both delivered at 10 m, so an NDVI map is a genuine neighbourhood-scale product - you can see a single large garden. On Landsat 8/9 (free, ~30 m) the same red and NIR bands give a coarser NDVI but one you can extend back decades for change work. India's Resourcesat LISS-III carries the same band pair for regional studies, and NDVI-type products are among the thematic layers on Bhuvan.

A caution that saves hours: band numbers differ between sensors (Sentinel-2's NIR is band 8, Landsat 8/9's is band 5). Always compute NDVI from the correctly identified red and NIR bands for your sensor, not from a remembered band number. And use surface-reflectance (atmospherically corrected) imagery when you can, especially for comparing dates.

NDVI is the workhorse, but it has known weak spots, and a family of indices patches them. Over sparse vegetation, bright or dark soil bleeds into the value; the Soil-Adjusted Vegetation Index (SAVI) adds a soil correction to reduce that. Over dense canopy NDVI flattens out (saturates); the Enhanced Vegetation Index (EVI) stays more responsive in lush conditions. There are indices tuned for water (NDWI) and built-up areas (NDBI) built on exactly the same normalised-difference trick with different band pairs.

For most architecture and planning work NDVI is enough - you want relative green cover, mapped consistently, not laboratory biomass. Reach for SAVI or EVI when your site is arid or heavily forested, and remember the golden rule: an index is a comparison tool. Its power is in differences - district to district, or the same place year to year - far more than in any single absolute value.