Normalized Difference Vegetation Index - Performance and Limitations

Performance and Limitations

It can be seen from its mathematical definition that the NDVI of an area containing a dense vegetation canopy will tend to positive values (say 0.3 to 0.8) while clouds and snow fields will be characterized by negative values of this index. Other targets on Earth visible from space include

• free standing water (e.g., oceans, seas, lakes and rivers) which have a rather low reflectance in both spectral bands (at least away from shores) and thus result in very low positive or even slightly negative NDVI values,

• soils which generally exhibit a near-infrared spectral reflectance somewhat larger than the red, and thus tend to also generate rather small positive NDVI values (say 0.1 to 0.2).

In addition to the simplicity of the algorithm and its capacity to broadly distinguish vegetated areas from other surface types, the NDVI also has the advantage of compressing the size of the data to be manipulated by a factor 2 (or more), since it replaces the two spectral bands by a single new field (eventually coded on 8 bits instead of the 10 or more bits of the original data).

Using the NDVI for quantitative assessments (as opposed to qualitative surveys as indicated above) raises a number of issues that may seriously limit the actual usefulness of this index if they are not properly addressed. Also, the NDVI has tended to be over-used (if not abused) in applications for which it was never designed. The following subsections review some of these issues.

• Mathematically, the sum and the difference of the two spectral channels contain the same information as the original data, but the difference alone (or the normalized difference) carries only part of the initial information. Whether the missing information is relevant or valuable is for the user to judge, but it is important to understand that an NDVI product carries only a fraction of the information available in the original spectral reflectance data.

• Users of NDVI have tended to estimate a large number of vegetation properties from the value of this index. Typical examples include the Leaf Area Index, biomass, chlorophyll concentration in leaves, plant productivity, fractional vegetation cover, accumulated rainfall, etc. Such relations are often derived by correlating space-derived NDVI values with ground-measured values of these variables. This approach raises further issues related to the spatial scale associated with the measurements, as satellite sensors always measure radiation quantities for areas substantially larger than those sampled by field instruments. Furthermore, it is of course illogical to claim that all these relations hold at once, because that would imply that all of these environmental properties would be directly and unequivocally related between themselves.

• The reflectance measurements should be relative to the same area and be acquired simultaneously. This may not be easy to achieve with instruments that acquire different spectral channels through different cameras or focal planes. Mis-registration of the spectral images may lead to substantial errors and unusable results.

Also, the calculation of the NDVI value turns out to be sensitive to a number of perturbing factors including

• Atmospheric effects: The actual composition of the atmosphere (in particular with respect to water vapor and aerosols) can significantly affect the measurements made in space. Hence, the latter may be misinterpreted if these effects are not properly taken into account (as is the case when the NDVI is calculated directly on the basis of raw measurements).

• Clouds: Deep (optically thick) clouds may be quite noticeable in satellite imagery and yield characteristic NDVI values that ease their screening. However, thin clouds (such as the ubiquitous cirrus), or small clouds with typical linear dimensions smaller than the diameter of the area actually sampled by the sensors, can significantly contaminate the measurements. Similarly, cloud shadows in areas that appear clear can affect NDVI values and lead to misinterpretations. These considerations are minimized by forming composite images from daily or near-daily images. Composite NDVI images have led to a large number of new vegetation applications where the NDVI or photosynthetic capacity varies over time.

• Soil effects: Soils tend to darken when wet, so that their reflectance is a direct function of water content. If the spectral response to moistening is not exactly the same in the two spectral bands, the NDVI of an area can appear to change as a result of soil moisture changes (precipitation or evaporation) and not because of vegetation changes.

• Anisotropic effects: All surfaces (whether natural or man-made) reflect light differently in different directions, and this form of anisotropy is generally spectrally dependent, even if the general tendency may be similar in these two spectral bands. As a result, the value of NDVI may depend on the particular anisotropy of the target and on the angular geometry of illumination and observation at the time of the measurements, and hence on the position of the target of interest within the swath of the instrument or the time of passage of the satellite over the site. This is particularly crucial in analyzing AVHRR data since the orbit of the NOAA platforms tended to drift in time. At the same time, the use of composite NDVI images minimizes these considerations and has led to global time series NDVI data sets spanning more than 25 years.

• Spectral effects: Since each sensor has its own characteristics and performances, in particular with respect to the position, width and shape of the spectral bands, a single formula like NDVI yields different results when applied to the measurements acquired by different instruments.

For these reasons, the NDVI should be used with great caution. 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.

A number of derivatives and alternatives to NDVI have been proposed in the scientific literature to address these limitations, including the Perpendicular Vegetation Index, the Soil-Adjusted Vegetation Index, the Atmospherically Resistant Vegetation Index and the Global Environment Monitoring Index. Each of these attempted to include intrinsic correction(s) for one or more perturbing factors. It is not until the mid-1990s, however, that a new generation of algorithms were proposed to estimate directly the biogeophysical variables of interest (e.g., the Fraction of Absorbed Photosynthetically Active Radiation or FAPAR), taking advantage of the enhanced performance and characteristics of modern sensors (in particular their multispectral and multiangular capabilities) to take all the perturbing factors into account. In spite of many possible perturbing factors upon the NDVI, it remains a valuable quantitative vegetation monitoring tool when the photosynthetic capacity of the land surface needs to be studied at the appropriate spatial scale for various phenomena.