Welcome to TERN Knowledge Base

Vegetation Indices based on Surface Spectral Reflectance and Absortion

Early sensors, such as NOAA’s AVHRR and NASA’s MSS, collected data in visible and NIR spectral bands. These sensors allowed the identification of areas of vegetation and investigation of vegetation state, based on the differential reflectance and absorption patterns of both bands in green plants. To do so multiple Vegetation Indices (VIs), combining in different ways estimates of the Earth’s surface reflectance in both bands, have been developed.

   Early VIs simply used the ratio of the estimates of reflectance for both bands; for example: red/NIR or NIR/Visible. However, these basic VIs have undesirable properties. For example, the latter index has a mathematical infinite range (from 0 to infinity). The most successful and better known of all VIs is the Normalised Difference Vegetation Index (NDVI, see below). The NDVI also has limitations, as it is sensitive to several perturbing factors. Consequently, multiple derivatives and alternatives addressing one or multiple of the NDVI limitations have proposed, such as the Enhanced Vegetation Index (EVI, also described below). Nevertheless, the NDVI remains a valuable and popular tool for vegetation monitoring.

    More recently, ‘modern sensors’ together with enhanced algorithms taking advantage of the improved performance and features of these sensors have permitted to directly estimate the biogeophysical variable of interest (see above). That is, to estimate FAPAR rather than the surface reflectance on the red and NIR bands.

Normalized Difference Vegetation Index (NDVI)

    The NDVI is the most successful and better-known VI. It uses RS measurements of the Earth’s surface spectral reflectance in the red and NIR bands to estimate the presence and condition of healthy green vegetation.

Formula & Range of Values

The NDVI formula is:

NDVI = (NIR – Red) / (NIR + Red)


NDVI = Normalized Difference Vegetation Index value.

NIR = Earth’s Surface Spectral Reflectance measurement in the Near Infra-red band.

Red = Earth’s Surface Spectral Reflectance measurement in the Red band.

    By design, NDVI values can range between +1.0 and -1.0. The interpretation of these and other typical values is a follows. The NDVI values below are approximations and ground truth points can/should be used to verify their values in the area of interest.

  • 0: It is obtained when NIR >> Red. This corresponds to the theoretical maximum value for healthy vegetation. In practice, this value is never reached. Tropical rainforest present display the highest NDVI values and can approach this value.
  • 3 - 0.8: Typical values for an area containing a healthy green vegetation canopy. Higher NDVI values correspond to higher densities of healthy green vegetation.
  • 2 - 0.4: Shrub and/or grassland. Tropical rainforest would have the highest NDVI values.
  • 2 - 0.5; Agriculture.
  • 1 - 0.2: Soils, as they generally exhibit NIR readings somewhat larger than Red reading.
  • Positive values close to 0: Possibly urbanized areas.
  • -1 – 0.1: Barren areas of rock and sand.
  • Very low positive to moderately negative values: Free-standing water (e.g. oceans, seas, lakes, rivers) and snow. However, depending on the fraction of snow cover NDVI can actually range between 0.75 and -0.15 (Dye and Tucker 2003).
  • Negative values: Atmospheric disturbance (e.g. clouds, water vapour, aerosols,…) and some specific materials.


NDVI has been widely used in numerous applications including (among others) agriculture, forestry, and environmental monitoring. In an agricultural setting NDVI has been used to measure crop biomass and precision farming. In forestry, NDVI has been used to predict forest supply and leaf area index (LAI). NDVI has found multiple uses in environmental monitoring, such as a drought indicator (water limits vegetation growth and density) and the assessment of ecosystem state.

    Time series of NDVIs can be used to monitor crop, forest and ecosystem dynamics of an area. To be able to compare NDVI over time (and space) is paramount that surface reflectance values are estimated from satellite images corrected for artefacts (e.g. atmospheric effects, illumination, a viewing geometry; see above).

    NDVIs has also been used for applications that this index was not originally designed for. For example because NDVI is directly related to photosynthetic capacity it has been used to estimate energy absorption of plant canopies. Similarly, NDVI has been used to estimate chlorophyll concentration in leaves, LAI, plant productivity, fractional vegetation cover, and accumulated rainfall. The rationale for these NDVI-derived estimations are typically observed correlations between NDVI values and ground-measured values across space. However, there are concerns about the suitability of these applications; for example, it is unclear the spatial scales of these associations.


   Unfortunately, it is now known that NDVI estimation is sensitive to several perturbing factors, including:

  • Atmospheric effects, such as those caused by water vapour and aerosols.
  • Cloud and cloud shadows. Clouds effects can be minimised by using composite images from daily or near-daily images.
  • Soil moisture effects: Soils tend to darken when wet, altering their reflectance in the Red and NIR bands. If changes in soil moisture affect differently the reflectance in each band, this can be mis-interpreted as changes in vegetation health or density.
  • Anisotropic effect: Surfaces reflect incoming solar radiation unevenly in different directions. Furthermore, anisotropic effect vary across the radiation spectrum. This effect can be important when the orbit of the RS platform drifts overtime (e.g. for NOAA’s AVHRR sensors). Composite images can also help to minimize this issue.
  • Sensor spectral effects: Because each sensor behaves in a particular way to the issues described above, different sensors (for different or even identical spectrum bands) will yield somehow different NDVI estimates.

   Therefore, NDVI estimates should be used with caution. Particularly in quantitative applications, the perturbing factors above described should be taken into account and when possible ameliorated.

   In any quantitative application that necessitates a given level of accuracy, all the perturbing factors that could result in errors or uncertainties of that order of magnitude should be explicitly taken into account; this may require extensive processing based on ancillary data and other sources of information. More recent versions of NDVI datasets have attempted to account for these complicating factors through processing.


   Because of the limitations of NDVI estimates described above, multiple alternative VIs that correct one or more of the perturbing factors affecting NDVI have been proposed. These include among others the Enhanced Vegetation Index (see below), the Perpendicular Vegetation Index, the Soil-Adjusted Vegetation Index, the Atmospherically Resistant Vegetation Index, and the Global Environment Monitoring Index. However despite its limitations, NDVI is a valuable tool when used correctly (i.e. minimising adequately the perturbing factors and at the appropriate spatial scale) that remains widely used.

Enhanced Vegetation Index (EVI)

   The EVI is VI developed to address some of the limitations of the NDVI. In particular, the EVI was designed to have improved sensitivity to changes in vegetation signal in areas of high biomass. This is a notable shortcoming of the NDVI. This is achieved by (1) reducing the effect of atmospheric artefacts in the VI, and (2) correcting for canopy background signals (Huete et al. 2002).

   The NDVI responds mostly to changes in the amount of chlorophyll present. The EVI, on the other hand, is more sensitive to canopy differences, such as canopy architecture/structure, canopy type, LAI, plant physiognomy, plant phenology, and stress. Therefore, both VIs complement each other, and in combination they can be used for an improved monitoring of vegetation dynamics and extraction of canopy biophysical parameters.

    Since the launch of the MODIS sensors on board the Terra and Aqua satellites, the EVI has been a standard product by NASA distributed by the Unites States Geological Survey (USGS). In this time, the EVI has become a highly popular product.

Formula & Range of Values

   EVI is computed as:

EVI = 2.5 * (NIR – Red) / (NIR + C1 * Red - C2 * Blue + L)


EVI = Enhanced Vegetation Index value.

NIR = Earth’s Surface Spectral Reflectance measurement in the Near Infra-red band.

Red = Earth’s Surface Spectral Reflectance measurement in the Red band.

Blue = Earth’s Surface Spectral Reflectance measurement in the Blue band.

C1 & C2 = Coefficients to correct for atmospheric condition (i.e. aerosols resistance). 

L = Canopy background correction.

   In the standard NASA MODIS EVI products, the values of these coefficients are C1 = 6, C2 = 7.5, and L =1. EVI’s values typically range from 0.0 to 1.0. A common EVI product consists of a single image layer, which cells values range from 0.0 to 1.0.

   A difference between NDVI and EVI is their behaviour in the presence of snow. In this case, NDVI decreases while EVI increases.


    The EVI is typically used at regional-scale to global-scale to assess the state and dynamics of plant biomass and biophysical properties (e.g. LAI). In combination with other parameters, such as Land Surface Temperature (LST), the EVI can be used to assess and map vegetation greenness-dryness and stress, as well as quantify evapotranspiration and water-use efficiency. A particular successful application of the EVI was the investigation of the Amazon forest seasonal growing patterns, which revealed a previously unsuspected distinct period of growth during the dry season (Heute et al. 2006). This pattern has significant implications for the Amazon carbon cycle and sinks, and consequently for the dynamics of greenhouse gases and global warming.

Limitations & Alternatives

       Limitation of the EVI include:

  • The signal to noise ratio of the blue band is quite poor, because the reflected energy over land in this band is extremely low.
  • EVI values cannot be compared to the rich historical measurements by AVHRR sensors, because its computation uses not only values for the NIR and Red bands, but also for the Blue band.

    The solution to both of these problems is the use of a new version of the EVI not including the Blue band. This is commonly referred to as a ‘2-band EVI’, with the acronym EVI2. That is, to compute a new type of EVI (EVI2) that is a function of only the NIR and Red bands, and minimises the difference between EVI and EVI2. There are infinite solutions to this problem, but imposing some conditions allows the estimation of the coefficients for a single solution. One of these solutions is:

EVI2 = 2.5 * ( (NIR-Red) / (NIR+2.4*Red+1) )


EVI2 = 2-band Enhanced Vegetation Index value.

NIR = Earth’s Surface Spectral Reflectance measurement in the Near Infra-red band.

Red = Earth’s Surface Spectral Reflectance measurement in the Red band.

   When the data quality is good and the atmospheric effects negligible, this formulation of the 2-band EVI (EVI2) has the best similarity with the ‘traditional’ 3-band EVI (EVI).

Fraction of Photosynthetic Radiation (fPAR)

   Photosynthetically active radiation (PAR) is the spectral range of the incoming solar electromagnetic radiation tat can be used by plants in photosynthesis. The PAR spectral region corresponds more or less with the range of light visible to the human eye (380nm-740nm, see above).

   The Fraction of Photosynthetically Active Radiation (fPAR), also known as the Fraction of Absorbed Photosynthetically Active Radiation (fAPAR), is the proportion of PAR that is absorbed by the canopy. fPAR is a parameter commonly used in RS and ecosystem modelling. It can be measured on the ground via handheld instruments or estimated from RS imagery at larger spatial scales.

   The fPAR can be speared into the fractions of PAR absorbed by the green (fPARgreen) and brown (fPARbrown) portions of the canopy. The fPARgreen can be further sub-divided into the fractions of PAR absorbed by the photosynthetic chlorophyll (fPARchl) and non-photosynthetic cellulose (fPARnon-chl). Mathematically this can be expressed as:

fPAR = fPARgreen + fPARbrown = fPARchl + fPARnon-chl + fPARbrown

Estimation & Range of Values

   There are three main methods to estimate fPAR values from RS images:

  • Linear Models (LMs): LMs use regression techniques to relate on-ground reflectance measurements and RS values. RS values can be surface reflectance readings or some derived VI (e.g. NDVI, EVI,…). For example, the ‘Chlorophyll Fraction of Photosynthetic Radiation’ MODIS product used the following relationship to estimate fPARchl: fPARchl = EVI x m + c (where: m = 1.112, c = -0.0746).
  • Physical & Biophysical Models: Physical models use physical principles of light (i.e. absorption and reflection rates at different surfaces) to estimate fPAR (and other parameters). Biophysical models also incorporate how light interact with the plants (e.g. photosynthesis, evapotranspiration, stress, and decay).
  • Artificial Neural Networks (ANN): ANN model non-linear and non-parametric systems to estimate fPAR using networks of typically simple algorithms.

   All of these methods require atmospheric correction and often use bidirectional reflectance normalization. Because the relationship between the fPAR estimates and satellite reflectance measurements vary with vegetation type, land cover type is also considered when estimating fPAR. Finally, composite images combining the best quality readings acquired over multiple days are used to minimise the presence of atmospheric artefacts (e.g. screening by clouds and/or snow).

   The final fPAR product is an image where cell values range between 0 and 1. Values close to 1 indicate that a high proportion of the PAR is absorbed by plants, while values close to 0 indicate little plant absorption of the PAR.


   Because vegetation growth is intrinsically related to the rate at which radiant energy is absorbed by plants, fPAR is an important parameter to estimate biomass production and vegetation health. Time series of fPAR estimates can be used classify vegetation (e.g. persistent and recurrent), as well as to monitor and analyse the long-term dynamics of vegetation cover.

   The fPAR RS estimates, together with other parameters (e.g. LST), can be used to compute surface photosynthesis, evapotranspiration, and net primary production. These parameters can be in turn used in ecosystem modelling to asses exchanges of water vapour, carbon dioxide, and energy between the Earth’s surface and the atmosphere.


  • Dye, D. G and Tucker, C. J. (2003). “Seasonality and trends of snow‐cover, vegetation index, and temperature in northern Eurasia”. Geophysical Research Letters, 30(7).
  • Huete, A. K. Didan, T. Miura, E. P. Rodriguez, X. Gao, L. G. Ferreira (2002). “Overview of the radiometric and biophysical performance of the MODIS vegetation indices”. Remote Sensing of Environment, 83.
  • Huete, A. R., K. Didan, Y. E. Shimabukuro, P. Ratana, S. R. Saleska, L. R. Hutyra, W. Yang, R. R. Nemani, and R. Myneni (2006). “Amazon rainforests green-up with sunlight in dry season”. Geophysical Research Letters, 33.

Provide your feedback about the experience with Knowledge base