NDVI
- Introduced by
- Rouse, Haas, Schell, and Deering
- Year
- 1973
- Type
- Spectral index for vegetation assessment
- Family
- Remote sensing vegetation monitoring
What does it do?
The Normalized Difference Vegetation Index (NDVI) is a spectral index computed from satellite or aerial multispectral imagery that quantifies vegetation greenness and vigor.
Introduced by Rouse and colleagues in 1973 using Landsat data, NDVI has become the most widely used remote sensing metric for vegetation monitoring, drought assessment, crop productivity forecasting, and land cover change detection.
The intuition behind the method
Vegetation reflects little visible (red) light but reflects a large fraction of near-infrared (NIR) light due to the structure and chlorophyll content of leaves. NDVI exploits this contrast: NDVI = (NIR - Red) / (NIR + Red), where NIR and Red are satellite-measured reflectance in those bands. NDVI ranges from -1 to +1: values near -1 indicate water (absorbs NIR), 0 indicates bare soil, and +0.5 to +0.9 indicates dense healthy vegetation. Time series of NDVI reveal seasonal cycles (vegetation greening in spring, senescence in fall) and inter-annual variability linked to water availability.
History
Introduced in 1973 using Landsat Multispectral Scanner data, NDVI rapidly became the standard vegetation index for agriculture, forestry, and drought monitoring. The launch of the Advanced Very High Resolution Radiometer (AVHRR) in 1981 and later MODIS (1999) enabled global daily NDVI monitoring. NOAA's Vegetation Health Index (VHI), combining NDVI and land surface temperature, became operational for drought early warning. Recent developments include fusion of high-resolution data (Planet, Maxar) with coarse-resolution data (MODIS) and deep learning for phenology detection.
When to use it
Use NDVI to monitor vegetation health at regional to global scales, assess drought impacts, forecast crop yields, and detect land cover changes. NDVI works best for large-scale assessments (pixel resolution >10 m) and is less useful for local-scale management where field-level detail is needed. Combine NDVI with other indices (EVI, SAVI) for robustness and with precipitation/temperature data for interpreting drivers.
Limitations and considerations
- Saturates at high vegetation density; two dense forests with different structure appear similar in NDVI
- Sensitive to atmospheric effects (aerosols, clouds) that introduce noise unless carefully corrected
- Cannot distinguish vegetation type or quality; healthy weeds appear similar to healthy crops
- Temporal resolution depends on cloud cover and satellite revisit frequency; gaps occur in cloudy regions
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Simple and computationally fast, requiring only two satellite bands widely available across sensors and archives
- Spatially continuous, enabling assessment across regions lacking field observations or ground infrastructure
- Strong empirical relationship to leaf area index (LAI), photosynthesis rates, and biomass
- Long temporal archive (Landsat since 1972, MODIS since 2000) enabling decadal trend analysis
Applications
- Drought early warning: monitoring NDVI anomalies to issue alerts for water stress before crop failure, guiding emergency response
- Crop yield prediction: correlating seasonal NDVI trajectories with final grain harvest to forecast production before season end
- Deforestation and land cover change detection: tracking NDVI decline to identify forest loss and characterize land use transitions
- Rangeland health assessment: using NDVI trends to evaluate overgrazing, recovery from drought, and carrying capacity in pastoral regions
How it works
How it works
Multispectral satellite imagery (Landsat, Sentinel, MODIS, Copernicus) with at least red and near-infrared bands is acquired. Data are radiometrically calibrated to at-sensor reflectance.
Atmospheric correction is applied to convert at-sensor reflectance to surface reflectance, accounting for aerosols, water vapor, and Rayleigh scattering. Cloud and cloud-shadow pixels are masked.
For each pixel, reflectance in the red band (0.6–0.7 µm) and near-infrared band (0.8–1.0 µm) are extracted.
NDVI is computed as (NIR - Red) / (NIR + Red) for each pixel. Time series stacks of NDVI are generated at regular intervals (daily for MODIS, 5–10 days for Landsat/Sentinel) to track seasonal and inter-annual variation.
NDVI maps reveal spatial patterns of vegetation abundance. Time series are analyzed for peak greenness, length of growing season, and drought-driven anomalies. Trends are correlated with precipitation, temperature, and land management changes.
Common pitfalls
- Misinterpreting spatial NDVI variation without accounting for soil background brightness, which affects NDVI in sparsely vegetated areas
- Comparing NDVI across different satellites or atmospheric correction algorithms without harmonization
- Over-applying NDVI to tropical regions where atmospheric aerosol abundance varies strongly, introducing systematic bias
Frequently asked questions
Why use NDVI instead of just looking at red reflectance?
Red reflectance alone doesn't account for soil background. A dark soil and a sparse plant canopy both have low red reflectance. By normalizing the difference between NIR and red by their sum, NDVI isolates the vegetation signal from soil. The resulting index correlates strongly with leaf area, photosynthesis, and biomass.
What are typical NDVI values for different land covers?
Water and snow: < 0.1; bare soil: 0.1–0.2; grass and shrubland: 0.2–0.5; crops (depending on growth stage): 0.3–0.8; dense forest: 0.6–0.9. Values vary seasonally (low in winter, peak in mid-summer in temperate regions).
How does cloud cover affect NDVI monitoring?
Clouds block the satellite view entirely; cloudy pixels are masked and must be interpolated or skipped. In tropical regions with persistent cloud cover, useful NDVI data may be infrequent (< 50% of observations clear). Radar-based indices (not affected by clouds) like NDVI-derived backscatter provide alternatives in cloudy areas.
Can NDVI distinguish between crop types?
Not directly. NDVI reflects greenness, not crop identity. Wheat and barley have similar NDVI trajectories. However, phenological timing (peak NDVI date, growing season length) can differ among crops, enabling crop classification when combined with weather data or field observations.
Related methods
Sources
Rouse, J. W., Haas, R. H., Schell, J. A., & Deering, D. W. (1973). Monitoring vegetation systems in the Great Plains with ERTS. Third Earth Resources Technology Satellite Symposium Proceedings, 1, 309-317.
Open publication referenceCopied.Jackson, R. D. (1983). Spectral indices in n-space. Remote Sensing of Environment, 13(5), 409-421.
Open DOI referenceCopied.