A lecture note on Analysis of reflectance spectra for identification of vegetation substance Redmond Ramin Shamshiri, PhD ramin.sh@ufl.edu

1. Page 1 of 2
Analysis of reflectance spectra for identification of vegetation substance, A Lecture Note. Redmond Ramin Shamshiri, Ph.D.
https://florida.academia.edu/RaminShamshiri
A lecture note on
Analysis of reflectance spectra for identification of vegetation substance
Redmond Ramin Shamshiri, PhD
ramin.sh@ufl.edu
Reflectance spectroscopy provides non-destructive, rapid,
reliable and precise analysis approach for plant’s health condition
assessment. Currently there are three main remote sensing platforms
for this purpose, including close-range platforms such as ground
based or handheld sensors, middle-range platforms such as drone-
based sensors or imaging devices mounted on autonomous
unmanned aerial vehicles (UAVs), and far-range platforms such as
piloted air planes or satellite-based sensors. Multi-spectral images
contain monochrome images taken from sensors with four or more
channels, each representing reflectance values for the objects in the
image corresponding to a specific wavelength. Hyperspectral images
can be acquired by ground-based, airborne or space borne
hyperspectral cameras, with each image pixel containing hundreds or
thousands wavelength bands (i.e., from 350 to 2500 nm). The spectral
data of a pixel in a hyperspectral image is a combination of the
reflected light from the entire features (i.e., moisture content and
chemical elements) that exist in that pixel area. For example, a typical
hyperspectral image in the size of 256 by 320 pixels that is acquired
with a 118-channels hyperspectral camera will have 81920 spectra
(or reflectance signal), each with a length of 118. The main difference
between hyperspectral imaging and spectroradiometer devices is that
data from hyperspectral camera includes spectral and spatial
information, whereas a spectroradiometer only provides reflectance
spectra. In the other words, a pixel in a hyperspectral image includes
reflected radiation for the piece of object in that pixel over hundreds
of wavelength bands. Indeed, the spectral data for that pixel has a
similar shape as data produced by a spectroradiometer. An image
from hyperspectral camera contains spatial axis, corresponding to a
line of pixels, and spectral axis, corresponding to reflectance data
distribution.
The idea behind analyzing of leaf reflectance spectra is to find a
correlation between the reflectance values from different leaves at a
specific wavelength, and the amount of nutrient concentration
obtained from laboratory results. Leaf reflectance spectra can be
collected in field using portable devices such as FieldSpec4 Hi-Res
Spectroradiometer (ASD, Inc., Boulder, Colorado), or PSR-3500 High
Resolution Full Range (350-2500nm) Portable Spectroradiometer
(Spectral evolution, Lawrence, MA) as shown in Figure 1. Spectral
data can be analyzed with different techniques such as wavelet
transformation for early detection of plants disease, water stress and
nutrient deficiencies, before visible symptoms can be identified.
Figure 1. Left: FieldSpec4 and right: PSR-3500, high resolution
Full Range (350-2500nm) Portable Spectroradiometer
To illustrate the concept of spectral analysis and how it can be
used for classification of a random agricultural sample based on a
specific physical or chemical property (such as color or nutrient
content), reflectance data of different samples that are generated
using full range PSR-3500 Spectroradiometer device are plotted
versus wavelength in Figure 2 A to D. To demonstrate the reflectance
differences between colors of a similar type material in the visible
range (visible wavelength lies between 380-700 nm, where different
colors are distinct for human eyes), reflectance spectra of four paper
samples with different colors (black, white, yellow and green) are
shown in Figure 2.A. An immediate observation is the distinct
differences between the black and white samples through the entire
wavelength band (350-2500 nm). This different is due to the fact that
black color paper sample absorbed almost the entire portion of light,
while white color paper sample reflected most of the intercepted
light, specifically in the visible range. On the same plot, the green
paper sample shows a pick value between 500-550 nm, which is
because of the wavelength range of green color that lies on 546 nm.
For lighter color paper samples, i.e., white and yellow, a sharp
increase in reflectance values between wavelengths of 350-500 nm
can be observed, followed by a decreasing trend that begins at around
1000 nm. Except for the black sample, reflectance values of other
three samples converge to 30% at 2500 nm.
Other than color properties, moisture content, thickness, and
cellular structure of leaflets are among other factors that influence
leaf reflectance spectra. Water stress and chlorophyll concentration
can be estimated directly from reflectance data. The source of
vegetation color in plants is due to the existence of chlorophyll that
absorbs blue and red wavelengths. That is the main reason that
healthy plants appear green in the visible range. To illustrate this,
spectral reflectance of a dark green citrus leaf and two sandy soil
samples, one wet and the other dry are provided in Figure 2.B. It can
be seen that the wet sand samples has a closer spectral shape to the
citrus leaf in the water band due to its moisture content, however the
dry sand sample has an entirely different spectral shape due to its
deficiency of moisture content. The difference between colors of sand
samples and citrus leaves samples with different green color due to
differences in their vegetation, as shown in Figure 2.C, can be referred
in visible wavelength range, where sand samples absorbed lesser
portion of light and reflected more because of having color lighter
than green. In the near infrared (wavelength range between 750-900
nm), the dry sand sample has significant deviation from the other two
samples. It can also be observed that none of the sand samples have
red edge (680-730 nm) when compared with the vegetation samples.
Water absorption band is approximately on 1400 and 1950 nm
wavelength and can be used to distinguish between real vegetation
and artificial vegetation (i.e., plastic leaf or green paper). Hence,
plant’s leaves suffering from water and nitrogen deficiency can be
distinguished with higher ratio of blue over red reflectance. It can be
clearly observed from Figure 2.B and Figure 2.C that beyond 1350 nm,
as wavelength increases, reflectance decreases, except in 1400 and
1900 nm (water absorption bands).
A more in-depth demonstration of water absorption band is
provided by plotting spectral signature for a green color artificial
plastic leaf, a green paper sample, and three real citrus leaves in
Figure 2.D. While the plastic leaf sample could be hardly distinguished
from the real leaves in visible range, they can be easily separated in
the NIR range (700-2500nm), specifically after 1300 nm. Plant’s
species type, age and its growth stage as well as possible stress and
deficiency affect reflectance measurements, especially in the NIR
range. Leaf reflectance is altered by stress more consistently at visible
wavelengths than in the reminder of the spectrum (730-2500 nm).
The difference in the spectral of the actual leaves sample is due to the
different levels of Chlorophyll concentration. This difference can also
be detected in the range of 400-700nm wavelength, or by referring to
the Nitrogen band which is approximately 550 nm. This is shown by
plotting spectral signature of five citrus leaves sample in Figure 3.A. A
closer view is provided in the range of 400-700nm in Figure 3.B
showing Nitrogen band and the differences between the five samples.
It can be seen that the first sample has the lowest level of Nitrogen
and has reflected more light in the Nitrogen band range. As the

2. Page 2 of 2
Analysis of reflectance spectra for identification of vegetation substance, A Lecture Note. Redmond Ramin Shamshiri, Ph.D.
https://florida.academia.edu/RaminShamshiri
amount of nitrogen increases in the leaves samples, the chlorophyll
content also increase in proportion and reflectance decreases. In fact,
a substantial amount of Nitrogen concentration is because of the
chlorophyll content. Nitrogen band is therefore used as an accurate
representation of plant vegetation status. Available commercial
cholorophyll meters that measures green color of leaflet sample are
not accurate due to their inconsistent readings caused by variability
in light source (shade, cloud, sunlight intensity). Other lower cost
instruments such as SPAD-502Plus uses red and NIR regions to
measure leaf absorbance, where absorbance is equal to
log(1/reflectance). SPAD calculates Ratio Vegetation Index (NIR/R), a
numerical value using R and NIR absorbance that is proportional to
the level of chlorophyll content in the leaf, and can be used for
assessing plant’s health condition. Another approach to distinguish
the differences between Chlorophyll concentration in samples is the
Normalized Difference Vegetation Index (NDVI), defined as
. Here, NIR is the near-infrared band
(740-760 nm) reflectance and R is the red band (640-620 nm)
reflectance. It can be implied from this index that higher NIR
reflectance and lower red reflectance values are resulted from yellow
or brown leaves. Therefore NDVI is a good indication for determine
growth stage (age), leaf area index and to some extend health state of
the plant. An example of such application includes identification of
infected and healthy wheat plots using NDVI. Other optical properties
of leaves around 700 nm and associations with stress detection,
chlorophyll content estimation, and physiological conditions of the
plant in blue-green (400-630 nm) and red to far red (630-800 nm)
can be addressed from correlating reflectance spectra with ground-
truth laboratory results.
Figure 2. Reflectance spectra of different substances, A: color paper, B:
sand and citrus leaf, C: different color citrus leaves, D: plastic leaf vs.
actual citrus leaves and a green paper
Figure 3. Differences in reflectance spectra of citrus leaves due to
different levels of Nitrogen
100 400 700 1000 1300 1600 1900 2200 2500
0
20
40
60
80
Wavelength (nm)
Reflectance(%)
GreenPaper
YellowPaper
WhitePaper
BlackPaper
A
100 400 700 1000 1300 1600 1900 2200 2500
-20
0
20
40
60
80
Wavelength (nm)
Reflectance(%)
WetSand
DrySand
DarkGreenLeaf
B
100 400 700 1000 1300 1600 1900 2200 2500
-10
0
10
20
30
40
50
Wavelength (nm)
Reflectance(%)
YellowGreenLeaf
LightGreenLeaf
DarkGreenLeaf
C
100 400 700 1000 1300 1600 1900 2200 2500
-30
-20
-10
0
10
20
30
40
50
60
Wavelength (nm)
Reflectance(%)
ArtificialLeaf
DarkGreenLeaf
LightGreenLeaf
YellowGreenLeaf
GreenPaper
D
400 600 800 1000 1200 1400 1600 1800 2000 2200 2500
0
10
20
30
40
50
60
Wavelength (nm)
Reflectance(%)
N5 (3.38%)
N4 (2.97%)
N3 (2.81%)
N2 (2.63%)
N1 (2.33%)
400 500 600 700
8
10
12
14
16
18
20
22
24
Wavelength (nm)
Reflectance(%)
N5 (3.38%)
N4 (2.97%)
N3 (2.81%)
N2 (2.63%)
N1 (2.33%)