International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
fields. The reflectance values for each of the bands analyzed
were taken from the image, through the pixel reader module.
The reflectance values used were the arithmetic average of the
individual values of all the pixels of each coffee field surveyed.
To obtain these values, the data was transformed to radiance
and then to reflectance, using the equations bellow:
VD,
E TA ( maxi — sii Y Laos
VD us
Where:
| = band;
VD; = digital value of the pixel, in band 1;
VD, ax = maximum digital value recorded
L4; 7 maximum radiance registered by the detector, in band 1;
Lmnini = Minimum radiance registered by the detector, in band 1;
mi
KR
BE, cosa
Where:
L = radiance;
D = distance Sun-Earth;
E = irradiance on top of the atmosphere;
[ = band;
a = solar zenital angle.
This data was transformed into percentage and attached to the
electronic sheets for statistical analysis.
Observing the nature of the data, which included a large number
of uncontrollable variables, obtained in field surveys, multiple
regression analysis were carried out, using models such as
stepwise and backward, scparating the data by farms, by
geomorphopedological environment and finally by production
region, Alto Paranaiba and Sul de Minas.
The analysis of the consistency of the database followed and
was carried out through analysis of frequency, mediums,
minimums and maximums. The results were satisfactory
enough to continue with the analytic procedures, carried out
through SAS (Statistical Analyses System) software.
The analyses of linear correlation of all the variables measured
on the field were carried out with reflectance measured in bands
3. 4 and 5. All variables were considered. The correlations
between the variables in question and the variables in band 4,
band responsible for the spectral response of the vegetation and
which correlated best with the variables, followed.
In a second stage, only the variables that presented the highest
correlation coefficients and/or the lowest significance levels
were considered. As the variable COV presented the best
correlations with all the other variables and with band 4 (level
of significance 0,0001/correlation coefficient 0,44753), it was
decided that coffee with ground cover under 50% and over 50%
would be correlated separately to homogenize the data. As the
spectral response depends heavily on the conditions of relief,
the data was analyzed in two groups: fields in slopes with more
than 15% gradients, and field in slopes under or equal to 15%
gradient. The consistency of this data was analyzed again,
followed by their correlations.
3. RESULTS AND DISCUSSION
3.1 Evaluation of the spectral response of coffee
The reflectance values of the coffee fields evaluated in band 3
are low, around 3%. This value was found for the plantations
with more than 50% of ground cover, as a result of one or a
combination of factors such as age of the crop, size of the
plants, spacing between rows, good vegetative state and vigour
among others. These characteristics refer to adult coffee plants
in good productive condition. The coffee still in formation
(under three years of age), which usually presented ground
cover lower than 50% lower presented higher reflectance values
in band 3, reaching 15%, as a result of the influence of the
substratum (soil and organic matter) on the spectral response.
This is due to the fact that in band 3, dense, green vegetation
presents great absorption, turning dark and allowing a good
contrast between the areas occupied by vegetation and those
unoccupied, such as exposed soil, roads and urban areas.
The reflectance results in band 4 are higher, reaching 35-40%
for formed coffee crops in good production condition. In the
coffee fields still in formation, again, given the great proportion
of exposed substratum, the reflectance values are lower (20-
25%). Band 4, which correspond to the near infrared (spectral
interval 0,76 to 0,90 pm), allow the dense, green, uniform
vegetation to reflect a lot of energy, appearing very bright.
Therefore, this band is the most recommended for the study of
spectral response in areas covered by vegetation.
The data analyzed shows that band 5 reflected more the
humidity of the soil and, indirectly, the type of soil, through its
greater or lesser capacity to retain water. Therefore, the
reflectance values were higher in the coffee crops in formation,
with great exposition of the substratum (up to 90%), reaching
values up to 28%. In the areas where the soil had greater water
holding capacity, such as the Latosols, the reflectance values
were lower, due to the absorption of water in this band.
3.2 Results of the statistical analysis
As the objective of the work was to analyze coffee areas and the
characteristics mentioned above, the analyses carried out for
band 4 will be presented.
Table 1 presents the results of the descriptive analysis of all the
data collected in the field, showing the crop variables evaluated
and their respective average reflectance values in band 4. This
analysis was carried to check the consistency of the data set and
assure their reliability for the following statistical analyses.
Linear statistic analysis was carried out, to cvaluate the linear
correlation coefficient and the level of significance of the
selected variables, as presented in Table 2.
The first analysis was carried out for all the data together. The
results show that among the variables analyzed, the one that
showed the lowest level of significance (0,0196) was COV.
despite the low correlation, which probably reflects the nature
of the data, as discussed before. This result is coherent with that
expected for band 4, since the variable COV (percentage of the
ground covered by the canopies of coffee plants in a coffee
field) is one that includes in its response the sum of the effects
of size, plant density, plants diameter, vegetative vigour and,
indirectly, the average production and/or productivity of that
field. It is possible to conclude, therefore, that it is the most
indicated variable to evaluate the spectral response of coffee
172
Internation
dI CHE IIT
fields in re
coverage o
response il
parameter )
better anal
carried out.
survey of a
Although t
significant
variable th
imaging, as
1978; Just
topographic
the spectra.
spectral res
the orientat
and positi
Therefore, |
statistical a
slope gradi
fields with
coffee fiel
separated fi
presented 1
(slope grac
(REFB4) w
than that «
gradient > 1
In the decli
as size, dia
levels of sig
The results
spectral res
influence it:
Statistical a
COV, whic
occupied by
Statist
|
Param
Average va
Standard d
Minimum +
Maximum
Table 1 -
Where:
REFB4: av
SIZE: avera
DENS: Plar
VIG: Vegete
DIAM: Diai
PROD: Pro
COV: Groui
SLO: Slope
