719
Wavelength (nm)
Figure 1: Parameters used in calculation of the Angular Vegetation Index (AVI)
The vegetation canopy reflectance was derived from a combination of the PROSPECT (Jacquemoud
and Baret 1990) leaf reflectance model and the SAIL canopy reflectance model (Verhoef 1984) with a
■bolt-on' hotspot (Jupp and Strahler 1991). The at-sensor radiance distribution function was generated
from the SAIL-generated bi-directional reflectance distribution function (BRDF) using the
LOWTRAN-7 atmospheric model (Kneizys et al. 1988). Finally this was convolved with the sensor
model to generate the at-sensor radiance received for the three channels of ATSR-2.
The relative sensitivity of the AVI was tested against the equivalent NDVI. In examination of the
sensitivity to the atmosphere the input parameters to PROSPECT were set to a healthy grass leaf (C,b
= 35 pg/cm 2 , C w = 0.01 cm and N = 1.5). Similarly the SAIL model was constrained to a dark soil,
spherical leaf angle distribution, and the leaf area index (LAI) varied from 0 to 6. View zenith and
azimuth were both 0* and solar zenith and azimuth set at 30’ and 30’ relative to North respectively.
The effect of atmosphere was tested by varying ground visibility from 10 to 50 km and the aerosol size
distribution using rural and maritime aerosol models provided in LOWTRAN-7. The effect of soil
colour was examined using typical 'light' (Caribou) and 'dark' (Berkshire) soils extracted from the
Purdue Soils Database, while the effect of leaf colour was tested by varying chlorophyll concentration
from 5 to 35 pg/cm 2 .
4. RESULTS
4.1 Sensitivity of AVI to Vegetation Fraction, Soil Brightness and Leaf Chlorophyll
Concentration
The AVI, calculated from canopy reflectance data, was tested for sensitivity to vegetation fraction
represented by LAI, which is used as the base variable for all the tests, soil brightness and leaf
chlorophyll concentration. In these tests and those involving the atmosphere the performance of AVI
was judged with respect to NDVI. The variation of AVI and NDVI with vegetation amount is shown in
Figure 2. Over a range of LAI of 0-6 both the indices are non-linear with NDVI in particular exhibiting
a steep increase in value from 0 to an asymptote at an LAI of 3. By contrast AVI continues to increase
to an LAI of 6. The effect of soil brightness on NDVI and AVI is revealed in Figure 3. The NDVI
exhibits a large decrease in value with increasing soil brightness in line with the observations of Huete
(1988)(Figure 3a). The AVI, however, is relatively insensitive to the effect of soil brightness when
computed for canopy reflectance data (Figure 3b). A difference in leaf chlorophyll concentration from
