1104 
2-VEGETATION INDICES 
The contrast in reflectance exhibited by green vegetation in the red and the near-infrared region can be 
exploited to identify the presence and possibly other characteristics of plants (Pinty et al., 1993). One such 
instrument is the Advanced Very High Resolution Radiometer (AVHRR) on-board the NOAA satellite series. 
In addition to the presence of two suitably channels, the avaibility of global data for the past ten years is a major 
advantage of this instrument for the study of long term changes in the environment. This paper will be 
concerned exclusively with the exploitation of these AVHRR data The design and use of analysis tools must 
be guided by specific applications. In particular, monitoring the state and the evolution of the vegetation cover 
may be entail a number of different requirements. At the most basic level, one may be interested in locating the 
great ecosystem types (e.g. forests, grasslands, deserts...). Information on the spatial distribution of vegetation 
is useful to quantify the extent and rate of deforestation, or to describe the overall structure of the environment, 
the location of ecotones, etc. Other applications require different information; one may be interested in 
estimating the amount of biomass, the exchange of water or carbon between the surface and the atmosphere, the 
productivity of an ecosystem, or the health of a plantation. 
2-1 Classical vegetation indices 
Three simple vegetation indices were initially designed to exploit the strong spectral contrast exhibited 
by green vegetation. The simplest such index is the Difference Vegetation Index (DVI) which is simply the 
difference between the near-infrared and the red reflectance of the surface. This index has not been used 
extensively, but the simple ratio (SR) defined as the ratio between near infrared and the red reflectance has 
been exploited by a number of authors. One of the advantages of this ratio form of this index is that it provides 
a partial normalization of some of the measurement perturbations. By far the most widely used simple index, 
however, is the Normalized Difference Vegetation Index (NDVI): 
although the Perpendicular Vegetation Index (PVI) was introduced by Richardson and Wiegand (1977) to 
better distinguish vegetated areas from bare soils. Strictly speaking, the higher these indices, the higher the 
probability that the area from which the measurements came is covered by green vegetation. 
2-2 Improv ed vegetation indices 
Early investigators noticed that these classical indices were quite sensitive to soil brightness changes. 
A variety of indices were then proposed to address this issue, including the Weighted Distance Vegetation 
Index (WDVI) 
WDVI = p n - a.p r 
of Clevers (1988) where a is the ratio of the near-infrared to red soil reflectance, and the Soil Adjusted 
Vegetation Index (SAVI) 
of Huete (1988). A single constant value of L = 0.5 has been found adequate in most situations. Later, Baret 
and Guyot (1991) proposed the Transformed Soil Adjusted Vegetation Index (TSAVI) based on soil line 
parameters with an adjustable X coefficient Qi et al. (1993) and Chehbouni et al. (1993) suggested to 
improve the SAVI by parameterizing the L itself as a function of the spectral observations. Their Modified 
Soil Adjusted Vegetation Index (MSAVI) incorporates a variable L = 1 - 2 x NDVI x WDVI, which is 
designed to automatically account for the varying fraction of bare soil sensed by the instrument as a function 
of the state of the vegetation. 
As the indices are computed on the basis of satellite data, atmospheric effects must also be taken into 
account As we don't know the vertical profile of various atmospheric constituents at the time of the satellite 
overpath, an alternative approach would be to incorporate a standard atmospheric correction into the definition 
of the index. This is the approach followed by Pinty and Verstraete (1992) in the definition of the Global 
Environmental Monitoring Index (GEMI): 
NDVI 
Pfl-Pr 
Pn+Pr 
SAVI = 
. Pr - 0.125 
GEMI = T) (1 - 0.25 II) - — 
where T) = 
2(Pn 2 -prV 1-5 Pn + 0-5 Pr 
Pn+Pr+0.5