The Normalized Difference Built-up Index (NDBI) highlights built-up land, such as roofs, roads and parking, in satellite imagery. Proposed by Zha, Gao and Ni in 2003, it's a quick way to map urban areas and watch towns grow over time.

How NDBI works

Many built surfaces, such as concrete, asphalt and some roofing, reflect more shortwave infrared (SWIR) than near-infrared (NIR). Vegetation does the opposite. NDBI compares the two:

NDBI = (SWIR1 − NIR) ÷ (SWIR1 + NIR)

Positive values suggest built-up land or bare ground; negative values suggest vegetation or water. It's the same two bands as the moisture index NDMI, with the sign flipped.

Where typical surfaces fall on the NDBI scaleNDBI runs from minus 1 to 1. Vegetation is about minus 0.34 and water minus 0.33. Bare soil is 0.12 and built-up land 0.17, close together, which is NDBI's main weakness.-1-0.500.51Vegetation −0.34Water −0.33Bare soil 0.12Built-up 0.17
Values use typical surface reflectance, the share of light reflected from 0 to 1. They're illustrative; real values vary with sensor, season and site.
Worked values
SurfaceNIRSWIR1NDBI
Vegetation0.450.22−0.34
Water0.020.01−0.33
Bare soil0.250.32+0.12
Built-up0.200.28+0.17

For built-up land: (0.28 − 0.20) ÷ (0.28 + 0.20) = 0.17.

NDBI's weak spot: bare soil

Look at the table again: bare soil (0.12) and built-up land (0.17) are very close. Dry fields, construction sites, sand and fallow farmland all score like towns. That's NDBI's main limitation, and why it's rarely used alone.

Common fixes:

  • Combine with NDVI. The "built-up index" BU = NDBI − NDVI (He and others, 2010) pushes vegetation far down, though soil and buildings still sit close together.
  • Mask water with MNDWI first, since water can also score oddly.
  • Use several dates. Fields change with the seasons; buildings don't.
  • Classify instead. Machine learning classifiers using all bands, plus building footprints where available, separate soil and buildings much better than any single index.

Which bands to use

Bands for NDBI
SatelliteNIRSWIR1Resolution
Sentinel-2B8 (or B8A)B1120 m for SWIR1
Landsat 8 and 9B5B630 m

At 20 to 30 m, one pixel covers several houses, gardens and street trees, so NDBI measures how built-up an area is rather than mapping single buildings. Landsat's archive back to the 1980s makes it good for tracking long-term growth.

Uses

  • Urban growth: how much land a town has built on since 2000, and where.
  • Impervious surface: estimating paved area for stormwater planning.
  • Urban heat: built-up land with little vegetation usually matches the hottest streets.
  • Planning baselines: a quick first map before detailed land use work.

Frequently asked questions

Can NDBI map individual buildings?

Not at 10 to 30 m. For buildings, use high-resolution imagery or existing building footprint datasets.

Why does my NDBI map show fields as urban?

Bare or dry soil reflects like built surfaces in SWIR and NIR. Use images from when fields are green, or combine with NDVI.

Is NDBI the same as impervious surface?

Not exactly. It's a proxy that correlates with built-up and impervious area, but needs calibration to estimate percentages.

Mapping urban growth or heat?

We'll map where your area has built up.

Built-up change, impervious surface and heat, with numbers by neighbourhood.

Talk to usSee heat maps