Verstraete algorithm, the Global Environment Monitoring Index (GEMI), is a non-linear index which
attempts to combine the sensitivity of the Simple Ratio (SR) for low vegetation cover and the NDVI
for deep vegetation while being resistant to atmospheric effects. Simulations with this index, however,
have shown it to have a small dynamic range and be highly sensitive to soil colour. The
Atmospherically Resistant Vegetation Index (ARVI) (Kaufman and Tanre 1992) takes advantage of
information provided by the blue waveband of the MODIS sensor to make NDVI less variable with
atmospheric perturbation. Unfortunately, this approach cannot be applied to ATSR-2 data, since the
assumptions behind the method cannot be translated to use a green waveband.
The focus of this preliminary work was to consider how the information from the three optical
channels of ATSR-2 could be used to provide a vegetation index that was resistant to atmospheric
scattering whilst at the same time being sensitive to variation in vegetation amount.
2. CONCEPT
Atmospheric scattering can be separated into molecular or Rayleigh scattering and aerosol or Mie
scattering. The former is characterised by a an optical thickness which decreases as a function of
wavelength approximately equal to A." 4 although the exact form is contingent on atmospheric pressure.
The effect of atmospheric aerosols on remote sensing is dependent on the chemical and physical
characteristics of the aerosol (Kaufman 1989), although if the aerosol is assumed to be spherical it can
be approximated by Mie Theory which also exhibits a decrease with wavelength. A further difficulty is
the spatial and temporal inhomogeneity of the aerosol loading. With reference to remote sensing of
vegetation the effect of atmospheric scattering declines with wavelength resulting in increased TOA
reflectance in visible wavelengths and when allied with continuum absorption a decline in TOA
reflectance in near-infrared wavelengths. Thus increased optical thickness can be equated to a
rotational movement with a large positive component in the green waveband, a smaller one in the red
and a negative component in the near-infrared. An index which is invariant with this rotational
movement should exhibit resistance to atmospheric effects.
The emergence of vegetation over a dark soil will result in an increase in reflectance in green
wavelengths, although as chlorophyll density increases this may decline again; a decrease in red
wavelengths associated with vegetation reflectance and a strong increase in near-infrared wavelengths.
Thus an index which represents the dynamic range of reflectance in each of these bands will provide a
good indication of vegetation presence. An index, the Angular Vegetation Index (AVI), which is
intended to incorporate the sensitivity to vegetation presence whilst being insensitive to the rotational
movement was defined using the three ATSR-2 channels to calculate the angle subtended in the red
chlorophyll absorption well (Figure 1):
AV I = tao -‘{i^k[p(A 3 )-p(A2)r 1 } + tan- 1 {iji.[p(i,)- /> (A 2 )f 1 } (1)
where p(Xj) is the radiance in band i normalised by the exo-atmospheric incident flux and Xj is the
centre wavelength of band i. The dependence on wavelength position is removed by normalising to
the centre wavelength, A-2- The value of AVI is scaled to 0-1 range by subtraction from 180 and
division by 90.
3. MATERIALS AND METHODS
The sensitivity of AVI to the presence of vegetation, atmospheric perturbation and soil colour was
tested using a forward model developed for simulating the at-sensor radiance for ATSR-2. The
simulator comprises scattering models for soil, leaves, vegetation canopy and atmosphere. Nadir
reflectance data obtained from the Purdue Soils database (Stoner et al. 1980) were used to represent the
soil. Off-nadir effects were generated by inversion of Hapke’s functions to derive the average particle
single scattering albedo, cd. The remaining coefficients were represented by typical values obtained
from other experiments.
