NDWI: Drawing the Line Between Land and Water
Water is one of the few surfaces that absorbs near-infrared almost completely, and NDWI turns that near-total absorption into a clean land/water boundary using just the green and NIR bands.
What it actually measures
McFeeters' NDWI maximises the contrast between water (high green, very low NIR) and everything else. Positive values flag open water while land, vegetation and soil fall negative. It was the first widely adopted single-index approach to surface-water delineation and remains a sensible default for lakes and large rivers.
How it is calculated
where Green is green reflectance, NIR is near-infrared reflectance.
What the numbers mean
> 0 open water · < 0 vegetation, soil or built-up. The land/water cut-off is often tuned slightly around 0 per scene.
Band configuration
| Sentinel-2 | Green B3 · NIR B8 |
| Landsat 8/9 | Green B3 · NIR B5 |
Where it is used
Use it for mapping reservoirs, lakes and wide river channels, and as a fast first pass for flood extent on clear scenes.
Limitations to know
Its well-known flaw is cities: built-up surfaces and deep shadow also suppress NIR and get misread as water. In any urban or mixed scene, MNDWI — which swaps NIR for SWIR — is the more reliable choice.
Compute NDWI on your own study area
Skip the code. Draw or upload a boundary and Spatial Research Suite runs NDWI on live Sentinel-2 or Landsat imagery — with cloud masking, exports and citations built in.
Run this analysis in GISforus →Frequently asked
What NDWI value counts as water?
Positive values (above roughly 0) usually indicate open water, but the exact cut-off is often tuned per scene.
Why does NDWI over-detect water in cities?
Built-up surfaces and shadow also have low NIR, so they can score positive; MNDWI corrects this by using SWIR instead of NIR.