Non-Homogeneous Feature Difference for urban applications

Used in urban mapping.

When to use

  • Built-up area mapping and urban extent monitoring
  • Impervious surface estimation for stormwater modelling
  • Urban heat island analysis (combined with thermal data)
  • Land cover change in expanding cities
  • Material classification within urban environments

Limitations

  • Material heterogeneity within urban pixels produces mixed signatures
  • Roof materials vary widely (metal, asphalt, tile, vegetation) within the same city
  • Shadow effects from tall buildings distort surface reflectance
  • Confusion with bare soil in arid environments is common
  • Temporal changes from construction require frequent updates

General Formula

RE1 700-710 nm

Sensor-Specific Formulas

Most-used sensors — click to show code below

SensorProviderFormulaBand Mapping
Wyvern(Band 16 - A) / (Band 16 + A)RE1→Band 16
ESA(B5 - A) / (B5 + A)RE1→B5
MAXAR(Red Edge - A) / (Red Edge + A)RE1→Red Edge
MAXAR(Red_Edge - A) / (Red_Edge + A)RE1→Red_Edge

Spectral Band Visualization — Dragonette-1

Code Examples

Adapted for Dragonette-1 bands —

nhfd_dragonette-001.py

Frequently Asked Questions

What is the NHFD (Non-Homogeneous Feature Difference) and when should I use it?

Non-Homogeneous Feature Difference for urban applications Urban and built-up indices distinguish impervious surfaces from natural land cover by leveraging the unique spectral properties of construction materials like concrete, asphalt, and metal roofing. NHFD is particularly suited for urban. The general formula is (RE1 - A) / (RE1 + A), which requires RE1 spectral bands.

Which satellite sensors can I use to calculate NHFD?

NHFD is supported by 16 satellite sensors in our database, including Dragonette-1, Dragonette-2/3, Gaofen-1, Gaofen-2, GeoEye-1 and 11 more. Each sensor uses different band designations — for example, Dragonette-1 uses the formula (Band 16 - A) / (Band 16 + A), while Dragonette-2/3 uses (Band20 - A) / (Band20 + A). Select a sensor above to see its specific band mapping.

What spectral bands does NHFD require and why?

NHFD requires RE1 (700-710 nm). These wavelength regions target the specific spectral features that this index is designed to measure.

How do I calculate NHFD in Python or R?

Both Python and R code samples are provided above. In Python, use rasterio to load individual band GeoTIFF files and numpy for the arithmetic. In R, the terra package handles raster operations efficiently. The key is to load bands as floating-point arrays to avoid integer division, and to handle division-by-zero cases where the denominator equals zero. For production use, consider applying a valid data mask to exclude no-data pixels before calculation.

NHFD vs other urban indices

IndexNameHow it differs
HHue IndexAlternative urban index — different band combination
IIntensity IndexAlternative urban index — different band combination
NDBINormalized Difference Built-up IndexAlternative urban index — different band combination
PISIPerpendicular Impervious Surface IndexAlternative urban index — different band combination

Related Urban Indices

References

https://www.semanticscholar.org/paper/Using-WorldView-2-Vis-NIR-MSI-Imagery-to-Support-Wolf/5e5063ccc4ee76b56b721c866e871d47a77f9fb4

Need help choosing?

Ask our AI assistant for sensor recommendations, code examples, or how NHFD compares to other indices for your specific use case.

Ask AI →