NDBI: Understanding the Built-up Index in Remote Sensing
The Normalized Difference Built-up Index (NDBI) is a very commonly used remote sensing index that distinguishes between built-up areas such as buildings, roads, concrete areas, roofs, and other non-permeable surfaces. This method distinguishes among urban locations, vegetation, water, and other land cover types by using the variations in the spectral responses of constructed surfaces in the near-infrared (NIR) spectrum and the shortwave infrared (SWIR) spectrum.
NDBI is one of the most important methods in urban mapping, land use and land cover classification, monitoring urban sprawl, analyzing impermeable surfaces, and planning in the area of urban remote sensing. NDBI can be calculated using multispectral satellite images and used in Geographic Information Systems (GIS) in conjunction with other spatial analysis processes.
What is NDBI?
NDBI is an acronym that stands for Normalized Difference Built-up Index. This index is used as a method of remote sensing to ease the process of detecting built-up land by making use of the properties of urban surfaces.
This index works by utilizing the light from the SWIR and NIR bands. The materials used in constructing a building have different spectral responses from plant materials as well as bodies of water; this allows the amount of built-up area to be quantified with the help of a mathematical formula.
NDBI usually uses the following equation:
NDBI = (SWIR − NIR) / (SWIR + NIR)
where:
SWIR = reflectance in the shortwave infrared band
NIR = reflectance in the near-infrared band
The normalized formulation produces values theoretically ranging from −1 to +1, although the actual range in satellite imagery depends on surface characteristics, atmospheric correction, sensor properties, and image processing.
How Does NDBI Work?
The NDBI concept has its roots in the spectral characteristics of different land cover systems.
Healthy vegetated areas necessarily have good reflectance qualities in the NIR region as a result of the internal structure of the leaf. At the same time, artificially created structures from concrete, asphalt, bricks, and roof tiles may demonstrate much higher reflectance values at certain SWIR wavelengths.
Having the nature of being the opposite of the optical principle behind some vegetation indices, NDBI’s value is positive whenever SWIR’s reflectance exceeds NIR’s reflectance. It also helps to spatially identify many human constructions.
For instance, if:
SWIR = 0.40
NIR = 0.20,
then:
NDBI = (0.40 − 0.20) / (0.40 + 0.20) = 0.33.
Thus, this means that in the case of positive NDBI values, SWIR reflectance is higher than the reflectance of electromagnetic waves in the visible spectrum due to the existence of artificially created constructions there. Still, it is important to mention that positive NDBI values should not be regarded as a kind of universal indication of human-made constructions.
NDBI Processing Workflow
The NDBI processing workflow is not limited to applying the formula to two raster bands.
Obtain Multispectral Data
Get multispectral data with bands covering the Near Infrared (NIR) and Shortwave Infrared (SWIR) wavelengths. The Landsat and Sentinel-2 satellites could be considered.
The data needs to have a high enough spatial resolution to match the scale of urban objects under study.
Apply Atmospheric Correction
For quantitative analysis, it would be better to use surface reflectance data or atmospherically corrected images rather than the original digital numbers.
Atmospheric correction helps to decrease the impact of atmospheric scattering and absorption on results.
Remove Clouds and Shadows
Clouds and their shadows cause abnormal responses in spectra. It is necessary to remove them from the images prior to NDBI calculation.
Match Spectral Bands
The SWIR and NIR datasets need to have the same:
Resolution
Coordinate reference system
Alignment of pixels
Extent of the rasters
Definition of NoData
For Sentinel-2, resampling might be required since relevant SWIR and NIR bands might have different intrinsic resolutions.
Calculate NDBI
Calculate the NDBI using the following normalized difference formula:
NDBI = (SWIR − NIR) / (SWIR + NIR)
NDBI can be automatically calculated using GIS software, raster calculators, Python, Google Earth Engine, and other remote sensing applications.
Analyze and Classify the Result
The resulting raster can be used in analysis alongside other indices and can be displayed with a continuous color scheme.
If researchers need to create a binary classification of built-up areas, they could set a certain threshold. However, this threshold needs to be set using local validation data.
How to Interpret NDBI Values
Since NDBI values are normalized, it is easy to interpret them to compare relative spectral differences.
Interpretation of NDBI may be summarized as follows:
Positive NDBI – Usually corresponds to impervious surfaces or built-up areas.
Value close to 0 – Usually corresponds to mixed or ambiguous spectra.
Negative NDBI – Usually corresponds to vegetation or water surfaces.
This, however, does not imply that this interpretation is rigid.
For example, bare soil and dry surfaces are likely to give positive NDBI values due to a similar spectrum to some of the built-up surfaces.
NDBI for Impervious Surface Mapping
Several impervious surfaces can be identified in the urban landscape, namely:
Buildings
Roads
Parking areas
Sidewalks
Concrete constructions
Industries
Airports
The information obtained from NDBI serves as an important initial step to identify these impervious surfaces. However, NDBI alone is not always capable of identifying all impervious materials.
For better classification, NDBI can be used along with:
MNDWI
Bare soil index
Land surface temperature
Texture features
Spectral classification
Machine learning algorithms
NDBI in GIS and Remote Sensing
NDBI rasters can be included in GIS workflows for spatial analysis.
Examples of GIS processes are:
Raster algebra
Classification
Zonal statistics
Overlay
Buffer
Change detection
Modeling urban growth
Area estimation
Comparison of administrative units
For example, one can overlay NDBI rasters with municipal borders to estimate the share of potentially built-up areas in each administrative unit.
Calculation of NDBI with Python
One can compute NDBI with the help of Python geoprocessing packages such as NumPy, Rasterio, and xarray.
Here is a basic implementation of NDBI calculation in the form of array operations:
import numpy as np
ndbi = (swir - nir) / (swir + nir)
However, in the case of real-world satellite imagery, one needs to account for the following issues:
NoData values
Zero division check
Raster metadata
Coordinate reference system
Alignment of pixels
Data type
Cloud masking
Scaling factors
A more robust implementation uses:
denominator = swir + nir
ndbi = np.where(
denominator != 0,
(swir - nir) / denominator,
np.nan
)
In case of processing large raster datasets, chunked operations with tools such as xarray and Dask can be considered.
Advantages of NDBI
There are several advantages of NDBI for urban remote sensing:
Mathematical simplicity
NDBI needs only two spectral bands.
Implementation is easy
NDBI can be computed via GIS raster calculator, Python, Google Earth Engine, and remote sensing software.
Effective for urban analysis
NDBI is able to identify lots of built-up and impervious areas.
Suitable for multispectral images
Spectral information from Landsat and Sentinel-2 could be used.
Suitable for multi-temporal applications
NDBI may help analyze urban development and land cover change.
Easily combined with other indices.
NDBI may be applied together with NDVI, NDWI, MNDWI, and thermal data.
Limitations of NDBI
However, there are some limitations related to NDBI despite its effectiveness.
Confusing Bare Soil with Built-up Surfaces
One of the major issues is bare soil, which may look like built-up areas and produce positive NDBI values.
Mixed Pixels
If the spatial resolution is not very high, then one pixel can include several land cover types such as a building, vegetation, and soil, thus leading to an intermediate NDBI value.
Material Properties Difference
Asphalt, concrete, metallic roofs, tiles, and other materials that are used in the construction process are characterized by different spectral signatures.
Effect of Vegetation and Surface Moisture
The condition of the vegetation and the moisture content of the surface might affect the NIR and SWIR reflectance and, therefore, the NDBI values.
NDBI Threshold Dependence
There is no universal NDBI threshold for any geographic location, season, sensor, and urban landscape.
Spatial Resolution
The detection of small buildings and thin roads depends greatly on the spatial resolution of the imagery.
The Normalized Difference Built-up Index (NDBI) is one of the key remote-sensing indices used to study built-up and urban surfaces. The core formula of NDBI
NDBI = (SWIR − NIR) / (SWIR + NIR)
employs the interrelation of SWIR and NIR spectra in order to enhance the importance of many built-up surfaces.
The application of NDBI can include urban expansion detection, impervious surface mapping, land use evaluation, urban planning, land cover classification, and urban environmental conditions research. However, positive values of NDBI do not necessarily mean buildings and urban structures. Bare soils, dried surfaces, mixed pixels, and specific features of the remote sensing tool can cause misclassifications.
For obtaining more accurate results, NDBI needs to be coupled with other indices (e.g., NDVI, MNDWI), correct pre-processing, multi-temporal analysis, and validation. Once implemented in GIS or cloud computing environments (Python), NDBI becomes a convenient and computationally cheap technique for obtaining data about the built environment from multispectral satellite imagery.
For more information or any questions regarding NDBI, please don't hesitate to contact us at
Email: info@geowgs84.com
USA (HQ): (720) 702–4849
(A GeoWGS84 Corp Company)
Comments