467 
« (o), 
imam 
* (o), 
trema 
püque 
lerosol 
" a 
ESTIMATION OF WINTER WHEAT GREEN LEAF AREA INDEX FROM FIELD SPECTROSCOPIC 
MEASUREMENTS USING A SEMI-DETERMINISTIC MODEL 
Robert DE WULF and Roland GOOSSENS 
Laboratory of Remote Sensing and Forest Management, Gent State University 
Coupure 653, B-9000 Gent, Belgium 
Georges HOFMAN 
Laboratory of Agricultural Soil Science, Gent State University 
Coupure 653, B-9000 Gent, Belgium 
ABSTRACT 
A comprehensive experiment to estimate green Leaf Area Index (LAI) of winter 
wheat (Triticum aestivum L.) from field spectroscopic measurements spanning 
four growing seasons (1984, 1985, 1986 and 1988) and involving three cultivara 
was conducted. 
Ground-based colour-infrared photography was the field spectroscopic method 
selected to acquire winter wheat canopy reflectance data. 
Seven types of 'vegetation index' were calculated. Correction procedures for 
solar zenith angle were introduced as experimental variable. 
Based on biophysical considerations, the monomolecular function was selected 
to model the relationship between vegetation indices and winter wheat LAI. 
Different models were constructed for pre-senescence and post-senescence data. 
Independent test data were used to assess the effectiveness of the prediction 
equations constructed from the training data. 
It is concluded that, despite obvious inaccuracies for large LAI values, the 
monomolecular function is an appropriate model to describe the relationship 
between winter wheat green LAI and vegetation indices. Ratio indices are 
better estimators of LAI than are orthogonal indices. For pre-senescence 
conditions, the Simple Ratio, Normalized Difference and TSAVI yield accurate 
and comparable results across cultivara and growing seasons. Post-senescence 
LAI is more difficult to estimate, and the Simple Ratio is the only valid 
vegetation index across cultiváis and seasons. A correction for solar zenith 
angle involving the normalisation of LAI yielded consistent good results. 
Key Words : Winter 
deterministic model. 
1. INTRODUCTION 
The use of leaf area as the description 
parameter in crop growth analysis was 
pioneered by Watson (1956) who defined 
it as "the area of leaf laminae per 
unit area of land surface ". 
The magnitude and duration of LAI is 
strongly related to the canopy's 
ability to intercept photosynthetically 
active radiation (PAR). Therefore, LAI 
is correlated with canopy 
photosynthesis and dry matter 
accumulation in situations where stress 
does not predominate. 
The importance of an accurate 
estimation of LAI for crop growth 
studies needs not to be stressed. 
Manual field methods involving cutting, 
sorting and weighing or planimetering 
are destructive and extremely tedious. 
In addition, in view of statistical 
considerations, they are not 
necessarily more accurate. For 
instance, Curran and Williamson (1985) 
showed that errors in ground data 
collection are likely to exceed the 
error in the remotely sensed data. 
Paradoxically, this would make remotely 
sensed data more accurate than the 
ground data used to check its accuracy. 
The rationale behind the use of 
multispectral reflectance data in 
wheat, Leaf Area Index, Vegetation Index, Semi- 
general, and of vegetation indices (VI) 
in particular, to estimate crop LAI 
has been established for some time. 
Canopy reflectance patterns in single 
bands lead to an explanation for the 
usefulness of Vi's as estimators of 
LAI. 
High absorption in the green and red 
wavebands causes rapid saturation in 
function of increasing LAI and occurs 
around LAI values of 2. 
On the other hand, near infrared (NIR) 
reflectance initially continues to 
increase at higher LAI levels due to 
multiple scattering between vegetative 
layers before eventually reaching an 
asymptotic level termed infinite 
reflectance, coinciding with a LAI 
value of about 8 (Wiegand et al. 1979) . 
From these considerations it follows 
that the relationships between LAI and 
Vi’s are basically non-linear. 
A multitude of empirical relationships 
between LAI and Vi's have been 
documented in literature. In view of a 
more operational use of these 
relationships, an investigation of a 
single model that would be valid across 
growing seasons and cultivars appears 
to be justified. For this type of 
exercise, field spectroscopy methods 
are eminently suited (Milton 1987). 
a»«® 
<4É¡ÍbÍS¡S 
«fett 
iM 
; I Ipi IWL, 
tPHII 
. >üüüi 
ïjpgl 
Ip 
» 
Imm