ChapterPDF Available

Abstract and Figures

Overview and basic concepts Remote sensing has been used in agriculture since 1970. The United States was the first country to use this technology in fields by collecting remotely sensed data to monitor wheat fields. Concomitantly, the first LANDSAT series satellite, was being launched to orbit providing images with higher spatial resolution and more frequently. Since then, the US became the leading country in research investments in this subject. In Brazil, the National Institute on Spatial Research (INPE) is responsible for the majority of the research and advances in terotechnology putting Brazil in the spotlight in the southern hemisphere. However, intense research in the use of geotechnologies in agriculture in Brazil only happened after there was a need for high food production using smaller areas. In current years, the concept of precision agriculture is widespread all over the country, and farmers are having more access to these geotechnologies. A straightforward definition of remote sensing is the capacity to obtain data from one or more objects through sensors without the need for direct contact. The information obtained is in terms of reflectance or emittance of electromagnetic energy. The sensors collect the amount of energy the objects reflect through photons radiation carried throughout the space in different wavelengths (Figure 1). The lower the frequency, the longer is the wavelength, and the lower is its energy (Heege, 2013). Each object responds differently in varying wavelengths of the electromagnetic spectrum, which in turn allows them to have a distinguishable spectral signature. Vegetation has a particular signature across the spectrum with overall lower reflectance in the visible region and a high reflectance on the near infra-red (NIR) which results in a steep increase between red and NIR bands. This narrow wavelength range between these two bands is known as red edge, and it will be discussed later in the chapter. Different plant components play significant roles in leaf reflectance in each band. For instance, the amount of light being reflected in the blue, green and red bands (0.4 µm a 0,7 µm) are heavily influenced by the pigments present in the leaves such as chlorophyll a, b and carotenoids. Together chlorophylls a and b absorb from 70 to 90% of all incident light. Nevertheless, chlorophyll "a" is responsible for absorbing 16 16. Remote sensing to predict peanut maturity 156 the majority of blue and red lights while absorbing very little green light. This higher reflectance of the green wavelengths can be observed across all pigments. Since green light is the most abundant light given off by the sun, plants developed an adaptive response to avoid tissue damage due to a high sensibility of pigments to greater energy levels. High energy can quickly destroy the pigments and become harmful for other plant structures and components such as DNA. In the NIR band (0.70 µm and 1.3 µm) there is a much lower pigment absorption than in the visible range. The reflectance in this band is mostly defined by the plant's internal structure, in particular, the spongy parenchyma mesophyll's internal air spaces. The air-water-cell interfaces within the parenchyma mesophyll have different refractive indices that will create a high scattering of NIR radiation, which in healthy plants it will translate to a reflectance of 40 to 60% of the radiation. In short wave infra-red (SWIR) (1,3 µm a 2,5 µm) the humidity inside the leaf is the primary factor influencing reflectance. The region of the spectrum ranging from 0.68 to 0.75 µm is one of the most extreme slopes formed in the signature curve of natural materials, and it is known as the red edge. The rapid change in reflectance observed in this region is characterized by the high pigment absorption in the red band and high scattering of NIR radiation. High chlorophyll content in the leaves will make the red edge shift towards longer wavelengths while lower chlorophyll content will shift the red edge towards...
Content may be subject to copyright.
Avanços na produção de amendoim
155
REMOTE SENSING TO PREDICT PEANUT MATURITY
Adão Felipe dos Santos
Lorena Nunes Lacerda
Rouverson Pereira da Silva
Overview and basic concepts
Remote sensing has been used in
agriculture since 1970. The United States
was the first country to use this technology
in fields by collecting remotely sensed data
to monitor wheat fields. Concomitantly, the
first LANDSAT series satellite, was being
launched to orbit providing images with
higher spatial resolution and more
frequently. Since then, the US became the
leading country in research investments in
this subject. In Brazil, the National Institute
on Spatial Research (INPE) is responsible
for the majority of the research and advances
in terotechnology putting Brazil in the
spotlight in the southern hemisphere.
However, intense research in the use of
geotechnologies in agriculture in Brazil only
happened after there was a need for high
food production using smaller areas. In
current years, the concept of precision
agriculture is widespread all over the
country, and farmers are having more access
to these geotechnologies.
A straightforward definition of remote
sensing is the capacity to obtain data from
one or more objects through sensors without
the need for direct contact. The information
obtained is in terms of reflectance or
emittance of electromagnetic energy. The
sensors collect the amount of energy the
objects reflect through photons radiation
carried throughout the space in different
wavelengths (Figure 1). The lower the
frequency, the longer is the wavelength, and
the lower is its energy (Heege, 2013). Each
object responds differently in varying
wavelengths of the electromagnetic
spectrum, which in turn allows them to have
a distinguishable spectral signature.
Vegetation has a particular signature across
the spectrum with overall lower reflectance
in the visible region and a high reflectance
on the near infra-red (NIR) which results in
a steep increase between red and NIR bands.
This narrow wavelength range between
these two bands is known as red edge, and it
will be discussed later in the chapter.
Different plant components play
significant roles in leaf reflectance in each
band. For instance, the amount of light being
reflected in the blue, green and red bands
(0.4 µm a 0,7 µm) are heavily influenced by
the pigments present in the leaves such as
chlorophyll a, b and carotenoids. Together
chlorophylls a and b absorb from 70 to 90%
of all incident light. Nevertheless,
chlorophyll "a" is responsible for absorbing
16
16. Remote sensing to predict peanut maturity
156
the majority of blue and red lights while
absorbing very little green light. This higher
reflectance of the green wavelengths can be
observed across all pigments. Since green
light is the most abundant light given off by
the sun, plants developed an adaptive
response to avoid tissue damage due to a
high sensibility of pigments to greater
energy levels. High energy can quickly
destroy the pigments and become harmful
for other plant structures and components
such as DNA.
In the NIR band (0.70 µm and 1.3 µm)
there is a much lower pigment absorption
than in the visible range. The reflectance in
this band is mostly defined by the plant's
internal structure, in particular, the spongy
parenchyma mesophyll's internal air spaces.
The air-water-cell interfaces within the
parenchyma mesophyll have different
refractive indices that will create a high
scattering of NIR radiation, which in healthy
plants it will translate to a reflectance of 40
to 60% of the radiation. In short wave infra-
red (SWIR) (1,3 µm a 2,5 µm) the humidity
inside the leaf is the primary factor
influencing reflectance.
The region of the spectrum ranging from
0.68 to 0.75 µm is one of the most extreme
slopes formed in the signature curve of
natural materials, and it is known as the red
edge. The rapid change in reflectance
observed in this region is characterized by
the high pigment absorption in the red band
and high scattering of NIR radiation. High
chlorophyll content in the leaves will make
the red edge shift towards longer
wavelengths while lower chlorophyll
content will shift the red edge towards
shorter wavelengths. The red-edge band
marks the transition from red to NIR. The
chlorophyll absorption in this range is much
lower than in the visible region, which
enables the red edge radiation to reach
deeper inside the canopy.
Figure 1. Signature curves of different objects in
the electromagnetic spectrum radiation.
(Elaborate by authors).
The remote sensing data during the crop
season may be useful in identifying areas
with the variability for detailed investigation
during the growing season using Vegetation
Indexes (VI’s). The VI's transform spectral
bands to a single variable and thus provide
information about the canopy of plants.
The Normalized Difference Vegetation
Index (NDVI) created by Rouse et al. (1973)
is the most known and widely used in
agriculture. Due it has a high correlation
with many agricultural features. Several
authors have used NDVI and found
correlation with leaf nitrogen content
(Amaral et al., 2015), plant height and
Avanços na produção de amendoim
157
number of branches per cotton plant (Souza
et al., 2017), yield at wheat (Marti et al. al.,
2007), and corn (Shanahan et al., 2001;
Oliveira et al., 2019). However, NDVI
offers limitations when the crop has a high
biomass condition, which makes the VI
saturate. Thus, it is essential to have
knowledge of the behavior between the IV
in the variable to be estimated remotely.
Several VI's are available and can be
correlated with different agricultural
features. However, to predict peanut
maturity has some divergence results among
the authors, some of them are showing the
next section.
Remote sensing in the peanut maturity
The use of remote sensing to estimate
maturity in crops such as soybean has led to
believe that remotely sensed data could be
an alternative. The more precise method of
estimating peanut maturity than the Hull
Scrap method (Willians and Drexler, 1981),
eliminating the subjectivity of the board
classification. Besides, satellite and UAV
images can provide a complete set of
information on maturity variability within a
field which is not feasible with the currently
used method. Studies have shown a good
correlation between vegetation indices and
peanut maturity. However, none of the
studies were able to select the best index.
Rowland et al. (2008) measured peanut
maturity with indices calculated from a
proximal terrestrial sensor that captures
reflected light in different wavelengths
(CropScan Multispectral Radiometer
CropScan Inc, Minnesota). The authors
found that the canopy reflectance in the
region of 830 to 850 nm showed a
significant correlation with crop maturity.
This result indicates that indices that use this
band are more prone to be better maturity
predictors and should be explored. A similar
study, Carley et al. (2008) used a portable
spectroradiometer (ASD FieldSpec Pro
Analytical Spectral Devices, Boulder, CO)
found that peanut maturity had a good
correlation with all bands used. When
analyzing the relationship between maturity
and vegetation indices, NDVI did not
correlate well when pods were mature.
Monsef et al. (2019) took a different
approach by capturing the plant's reflectance
with a UAV-mounted camera and
subsequently using the data to build a model
to predict the maturity. Although the model
did not show good accuracy at 60 days after
planting, the accuracy increased as plants
stayed in the field. The authors attributed
this fact to a difference in chlorophyll “a”
and “b” since it became more prominent as
the plants mature. Based on these results, it
is suggested that data should be collected
starting two months after planting, and
frequency of data collection must be
increased as crop approaches harvest. The
study did not include an assessment of the
16. Remote sensing to predict peanut maturity
158
chlorophyll “a” and “b” difference
evidencing that there is a gap in knowledge
of the physiological processes that occur in
the peanut plant during pod maturity.
Santos et al. (2019) tested different
vegetation indices to estimate peanut pod
maturity variability. In this study, using a
multispectral camera, numerous vegetation
indices were calculated. It was found that
indices containing the red edge and NIR
bands had the highest performance in
predicting maturity than indices using bands
in the visible range (Figure 2). This
performance difference can be attributed to
the saturation problem on indices calculated
using visible bands such as NDVI, GNDVI.
Although promising, further research is
needed to check the stability of indices
performance in different peanut cultivars
and growing seasons.
(a)
(b)
Figure 2. Modified vegetation indices NLIre (A) MNLIre (B). (unpublished data)
y = -0.8926x + 0.2892
R² = 0.611
RMSE= 0.154
y = -2.14x + 0.1152
R² = 0.671
RMSE= 0.143
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
-0.55 -0.45 -0.35 -0.25 -0.15 -0.05 0.05 0.15 0.25
PMI
NLI (730-740)
Irrigation Dryland
Irrigated
Avanços na produção de amendoim
159
Perspective to future, challenges and final
considerations
With the rapid increase in agriculture
technology, in the future, the peanut growers
across the globe could use the remote
sensing in your farms. Especially to
management the diseases, predict maturity
and precision harvest. However, the
researchers still some challenges to solve.
All of the results prove that it is possible
to use remote sensing to predict peanut
maturity, especially using a multispectral
camera and satellite images. These results
allow us to advance the knowledge to search
for a non-destructive method to estimate
peanut maturity. However, despite the
satisfactory results found, more years of
study are needed with remote sensing to
predict peanut maturity. Moreover, need to
find physiological relationships between the
peanut plant, especially leaf components,
changes with peanut maturity and then
create one estimate the days until harvest
based at vegetation indices.
In this chapter, we only mention NDVI at
the first part, since so far this is the most
studied, the first results using other IV are
from our research group and have not yet
published. However, it is there are many
other vegetation indices described at
literature.
16. Remote sensing to predict peanut maturity
160
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.