During the 1982 growing season, reflectance mea
surements of bare soil were obtained with a field
spectroradiometer (described by Clevers, 1986c).
Results are illustrated in figure 2 for a wet and
for a dry soil. Estimated reflectances for the (si
mulated) spectral bands used with multispectral
aerial photography (MSP) are listed in table 1.
These results confirm the validity of the assump
tion that the ratio between the reflectance in any
pair of the green, red and infrared bands is con
stant for the soil type investigated. These ratios
also tend to the value one.
REFLECTANCE GREEN RED INFRARED
Figure 2: Reflectance of bare soil as measured by
means of the field spectroradiometer on two dates
during the 1982 growing season.
: wet soil (28 April 1982)
: dry soil (10 June 1982).
Table 1: Estimated reflectances of a wet soil and a
dry soil for the MSP bands, ascertained by means
of the field spectroradiometer.
reflectance (%)
green red infrared green/red infrared/
red
28 April 1982 12.4 13.6 15.0 0.91 1.10
10 June 1982 20.5 23.7 26.3 0.86 1.11
4.5 Estimating leaf area index
The dates of the flights and of the field sampling
did not coincide and therefore the smoothed LAI da
ta provided the possibility of interpolating the LAI
values for the dates of flight missions. These data
were used for studying the relationship with reflec
tance measurements.
Method 0 for estimating LAI follows the line that
the reflectance of bare soil and the reflectance of
vegetation in the different spectral bands is known.
For instance, after estimating soil cover by means
of equation (2), the corrected infrared reflectance
can be calculated by means of equation (3). Finally,
LAI can be estimated by means of equation (5). The
reflectance of vegetation may vary slightly because
of differences in leaf colour, but the mean value
(averaged over different dates and treatments) was
used for the green and red bands: r = 5.0 and
r = 2.0 (see Clevers, 1986c). During the 1982
seAson the soil was dry during missions up to mid-
June (complete soil cover with barley), because
there was little rainfall on days prior to the mis
sions. Hence the reflectance value for bare soil in
each band could be considered as being approximately
constant. The following reflectance values for bare
soil were used (see Clevers, 1986c):
r = 12.4, r = 13.5 and r s ^ = 14-7.
Mltflod 1 uses equation (4) for S 6aiculating the cor
rected infrared reflectance. For Cj a value of 0.9
and for C 2 a value of 1.1 was applied. The reflec
tances of the soil are not needed explicitly.
Method 2 concerns the crude approximation of the
corrected infrared reflectance as described by equa
tion (8). This may give a good approximation of the
corrected infrared reflectance, since the ratio of
green reflectance to red reflectance and of infra
red reflectance to red reflectance of the soil at
the experimental farm were nearly equal to one. The
reflectance of bare soil and of vegetation in the
different spectral bands are not needed explicitly.
Theoretically, method 1 is considered to be a more
accurate method of correcting the infrared reflec
tance for background than method 2 since the actual
ratios of bare soil reflectance in the various spec
tral bands are used. A disadvantage of method 1 com
pared with method 2 is the fact that measurements in
three spectral bands are required (of which the green
and red are mostly strongly correlated) for calcu
lating the corrected infrared reflectance, whereas
with method 2 only the red and infrared reflectances
are required. These latter two methods (1 and 2) may
be superior to method 0 if soil reflectance cannot
be regarded to remain constant between the recording
dates, because neither of these methods explicitly
require the reflectance values of soil. Results for
all three methods are presented in table 2 and fi
gures 3, 4 and 5.
Table 2 shows that both method 0 and 2 gave sim
ilar results. Method 1 gave slightly worse results
(larger coefficients of variation). This indicates
that method 2 may be very useful if soil reflec
tance is not similar on different dates. The con
clusion that the simplifications induced by method 2
did not yield worse results than method 0 or 1 was
caused by the fact that the estimation of LAI is
empirical. For instance, f M . is the asymptotic
value of the corrected infrafea reflectance, and it
is estimated from the measurements (e.g. from a
training set). For most data sets, r . is an ex
trapolation of the data, and this resiífís in consid
erable variation in estimated r . values, but with
oo xj-
an optimal fit to the data. '
An important conclusion derived from the regres
sion of LAI on corrected infrared reflectance for
the vegetative stage of barley was that measure
ments of all dates may be combined, resulting in
one curve. So, leaf angle distribution did not vary
to such an extent as to disturb the fit of the curve.
Similar results were found for other cereals and in
other seasons (Clevers, 1986c).
A very important analysis of field measurements
from agricultural field trials is an analysis of
variance in order to investigate whether treatment
effects are significant and whether interactions
occur. Analogously with field measurements (LAI),
an analysis of variance can be carried out on re
flectance measurements in the various spectral
bands for investigating whether the latter vari
ables can be ascertained with relatively smaller
variance and whether treatment effects can be as
certained with larger power than by means of con
ventional field measurements. An analysis of vari
ance can also be carried out on the estimated LAI
from the reflectance measurements. The results of
this latter analysis are more comparable with the
analysis of variance applied to the original LAI
measurements, since they involve the same variable.
Table 2: Results of different methods for correcting
the infrared reflectance and subsequently estimating
LAI. Field trial 116 in 1982. (CV = coefficient of
variation).
method
equation
number
â
r
»,ir
CV of
residuals
figure
0
2 and 3
0.255
71.21
0.217
3
1
4
0.228
75.70
0.230
4
2
8
0.252
68.57
0.214
5