diversity of the major sugar beet growing areas.
The predictive value of the model is rather poor
considering the fact that pluviometric data till
October 31st are to be included. In the model the
leaf yellowing percentage is the only non-
meteorological factor and has to be estimated
visually at the end of August. Leaf chlorosis can be
detected successfully by multispectral remote sensing
(Andrieu 1977, Boehnel et al 1983, Reichert 1983).
In practice the Royal Belgian Institute for Sugar
Beet Amelioration relies on biweekly sampling in
the major sugar beet growing areas to assess biomass
and formulate trends.
3 MATERIALS AND METHODS
3.1. Collection of agronomic data
The experimental results reported herein were
obtained in 1984 and 1985 from test fields located
in Sint-Laureins, in the north-west of Flanders. Soil
type can be characterized as an aquic udifluvent.
The area constitutes one of the richest agricul
tural regions featuring crops as winter wheat, winter
barley, sugar beets and potatoes.
Sugar beets cv "Allyx" were sown on 23.04.1984 and on
22.04.1985 respectively. Emergence was recorded on
14.04.1984 and on 05.05,1985. Application of
fertilizer and herbicides was similar during both
seasons and is not further discussed here. Traces of
fungicide sulphur spray could be detected on leaves
on 21.08.1984. The use of insecticides on other dates
did not leave visible traces on the canopy.
Originally a 10 day interval sampling scheme was
envisaged. Experiments were to be conducted only when
a direct solar path free of clouds was available.
Adverse weather conditions occurred frequently during
both growing seasons. This resulted in 10 sampling
dates in 1984 and 9 sampling dates in 1985.
On each sampling date 3 plots were located by
throwing a 80cm by 80cm frame onto the canopy from
randomly defined coordinates in the test field. All
plants inside the frame were extracted. Green leaves
were separated from the roots, spread out on a matt-
black painted surface and recorded on black-and-white
film. All roots (with heads cut off) were weighed
immediately to determine fresh weight. Three randomly
selected roots per plot were retained for sugar
content determination and deep-frozen to avoid sugar
loss awaiting analysis.
Besides root fresh weight and sugar content (%) green
leaf area index (GLAI) was determined. The latter was
accomplished by placing a transparent dot grid (4mm
spacings) on top of 18cm x 24cm sized photographic
paper during enlargement. By counting the dots on the
leaves GLAI could be computed as dimensions of black
background and sample plot were known. This method
proved to be much faster than the use of a polar
planimeter. Parallel measurements using both methods
on randomly selected samples did not reveal
significant differences (Student's T test).
3.2. Extraction of multispectral data
Simultaneous recording of the test plots on CIR film
assured direct relationship of biomass to
multispectral parameters to be extracted from the
images. A MIRANDA SENSOREX EE (*) camera fitted with
standard 50 mm optics was mounted on a light weight
metal boom supported by a ZEISS tripod. The camera
shutter could be triggered using an air release.A B&W
No Y3 filter, bearing close resemblance to the widely
used VRATTEN 12 filter, was used throughout the
(*) This and other trademarks are mentioned for
information only. No endorsement by the authors is
implied.
experiments. The selected emulsion was a 35mm KODAK
EKTACHROME 2236 film.
Pictures were taken at solar noon plus or minus 2
hours to avoid influences by changing solar zenith
angle. The plane of the film was kept parallel to the
plane of the canopy. Camera distance to ground level
was fixed at 2.30 metres. An area of lm by 1.5m
filled the 24 x 36 mm frame. A KODAK grey card with
known reflectance characteristics was included in
every picture. A second reference panel was judged
unnecessary as the amount of airlight between camera
and target was considered negligible in relation to
total reflected energy. Conventional light meters do
not indicate proper diaphragm/shutter speed
combinations for infrared-sensitive emulsions. Hence
series of 4 slightly different combinations were made
using an approximate meter setting for 50 ASA, It was
experienced that at least one slide was suitable for
subsequent processing.
Exposed emulsions were developed according to NASA
standards. From each development batch a strip of
film was exposed through a step wedge in a
sensitometer. The step wedge image was used for
preparation of characteristic curves.
On each slide 50 randomly located points were
measured using a MACBETH TD 504 transmission densito
meter. The filter assembly consisted of VRATTEN No 94
(blue), No 93 (green) and No 92 (red) filters
allowing measurements of integral densities relating
to green, red and infrared reflectance respectively.
A .5 mm aperture was used. Three measurements were
made on the reference panel. Integral densities were
converted to analytical densities using unit spectral
dye density data for the CIR emulsion. Green, red and
infrared reflectance, as well as a number of
vegetation indices and transformations were
calculated for each slide.
Brightness, Greenness and Yellowness were determined
using the method outlined by Jackson (1982).
Following linear combinations similar to Tasseled Cap
transformations for Landsat data were obtained:
Brightness= .603*G + .489*R + .629*IR
Greenness =-.292tG - .598*R + .746*IR
Yellowness=-.794*G + .634*R + .217*IR
Eventually only Greenness was retained for further
analysis due to its proven relationship to green
biomass. In addition to the IR/R ratio (Vegetation
Index or VI) and Normalized Difference or ND
(SQRT((IR-R)/(IR+R) + .5)), the Perpendicular
Vegetation Index (PVI) as the distance to the Soil
Line in IR/R space (Richardson and Viegand 1977) was
calculated. Figure 1 shows the soil lines for the
test fields.Full details on methodology, applied
software and hardware configuration used for
extraction of reflectance from CIR slides can be
found in De Vulf and Goossens (under review).
Cover percentage could be calculated by placing a dot
grid on projected slides and counting the amount of
points falling on vegetation.
Figure 1. Soil line for an aquic udifluvent.
4 RESULTS
4,1. Gree
Developme
both grovi
significa
featured
stronger
measured
authors f
51 8
ÜÜ Q
a
if
a S
“W-
Figure 2.
GLAI (dai
and 1985
Linear re
been repc
Badhwar e
Asymptoti
combinati
GLAI vali
agreement
to satur;
be est i me
This put
procedure
higher GI
Figure 3
sugar bee
Best fit
and speci
The indi
predictec
and plot
(Figure ■
for high
powerful
should
coefficie
power.
