The calculation of the indices 1 to 4 
is straightforward and they are 
site-independent. 
The RIV and NDIV indices are amendments 
of the well-known SR and ND. Their 
construction was proposed to 
investigate the information value of 
the green waveband in a ratio index. 
This choice is also to some extent 
influenced by the work of Sellers 
(1989), who suggested that ideally a 
ratio VI should be close to 2. Hence 
the replacement of R by the average 
reflectance in the PAR wavelenghts 
could yield values closer to 2. 
In the formulae of SR and ND, R has 
been replaced by (G+R)/2, which is the 
closest approximation of PAR 
reflectance with the data available. 
The other indices included in this 
study, PVI, TSAVI and Tasseled Cap-like 
transformations are site-dependent, as 
they include bare soil reflectance 
components . 
Calculation of PVI and TSAVI require 
knowledge of the R and IR values for 
bare soil. 
R-IR data sets of 
prepared by random 
available CIR 
Eventually 100 R-IR 
obtained which allowed for the 
construction of the soil line of the 
test area. All soil moisture conditions 
were covered in order to yield a valid 
soil line. 
bare soil were 
sampling of all 
transparencies. 
data pairs were 
5. Perpendicular Vegetation Index (PVI) 
(Richardson and Wiegand 1977) 
PVI = (-1.134 x R - IR - .674) / 1.319 
6. Transformed Soil-Adjusted Vegetation 
Index (TSAVI) (Baret et al . 1989) 
1.134 x (IR - 1.134 x R - .674) 
TSAVI = 
(R + IR x 1.134 - .764) 
Tasseled Cap-like transformations were 
obtained according to a procedure for 
calculation of coefficients of n-space 
indices as proposed by Jackson (1983). 
Only the linear combination 
corresponding to the Tasseled Cap 
'greenness' (Kauth and Thomas, cit. 
Bariou et al. 1985) was retained. 
7. Greenness (GRS) 
GRS = -.183 * G - .723 * R + .665 * IR 
In addition to the investigation of 
several VI types, four types of 
normalisation of the spectral data for 
the effect of solar zenith angle (G) 
were considered: 
1. No correction, i.e. assuming that 
the precaution of taking field 
spectroscopic measurements close to 
solar noon is sufficient ('nocor'). 
2. Blanket normalisation of all 
vegetation indices by multiplication 
with cos (G). This method is equivalent 
to data normalization to a standard sun 
elevation, such as implemented by 
Pinter et al. (1983) and Tueller and 
Oleson (1989) ('bcor'). 
3. Normalisation with cos (0) of 
visible wavelength bands only, 
considering experimental evidence that 
IR reflectance is usually less 
influenced by solar zenith angle than 
reflectance in the visible wavebands, 
especially in the case of incomplete 
canopies (e.g. Ranson et al. 1985) 
('vcor') 
4. Normalisation of LAI with the cosine 
of the solar zenith angle as proposed 
by Wiegand and Hatfield (1988). Hence, 
in contrast to alternatives 2. and 3., 
the causal variable in the LAI/VI 
relationship is normalized ('lcor'). 
3. MODELLING 
3.1. Model construction 
The fact that different authors propose 
different LAI/VI models for the same 
crop leads to the assumption that 
probably no unique index for all 
locations or crops exist. A possible 
explanation can be found in the use of 
different field spectroscopy 
instrumentation and the use of 
different wavebands to construct a VI. 
In addition, different sampling 
procedures are prone to seriously 
influence calibration relationships. 
Linear empirical models have been 
suggested to characterize the 
relationship between winter wheat LAI 
and Vi's (e.g. Hinzman et al. 1986, 
Major et al. 1986). 
However these authors based their 
models on LAI data generally lower 
than 3. Still, a linear model was found 
to provide better estimates of spring 
wheat LAI than a leaf area simulation 
model (Kanemasu et al. 1985). 
However, in search of a model type 
describing the LAI-VI relationship, the 
asymptotic behaviour between both 
parameters cannot be ignored on 
theoretical and experimental grounds 
(e.g. Asrar et al. 1985). 
Earlier proposed non-linear statistical 
models include exponential regressions 
(Hinzman et al. 1986) and second 
degree polynomials (Bauer et al. 1981) 
In addition, Asrar et al. (1985) and 
Hatfield et al. (1985) showed that 
empirical relations were different for 
the pre- and post-senescence period. A 
hysteresis-like functional relationship 
should be assumed for the complete 
growth cycle of winter wheat. This can 
be explained by the fact that 
radiation phenomena are different for 
green and senescing canopies. 
Adoption of a semi-deterministic model 
can only be justified if a physical 
relationship between a VI and LAI 
exists. 
An appropriate choice seems to be the 
monomolecular curve (Hunt 1982) 
formally expressed as 
469