Available via license: CC BY 4.0
Content may be subject to copyright.
Sensors 2023, 23, 9562. https://doi.org/10.3390/s23239562 www.mdpi.com/journal/sensors
Article
Analysis of Cadmium Contamination in Leuce
(Lactuca sativa L.) Using Visible-Near Infrared
Reectance Spectroscopy
Lina Zhou 1, Leijinyu Zhou 1, Hongbo Wu 1, Lijuan Kong 1, Jinsheng Li 1, Jianlei Qiao 2 and Limei Chen 1,*
1 College of Engineering and Technology, Jilin Agricultural University, Changchun 130118, China;
zhoulina976430@163.com (L.Z.); zhouleijinyu8514@163.com (L.Z.); whb05082023@163.com (H.W.);
konglijuan630@sina.com (L.K.); jinshengl@jlau.edu.cn (J.L.)
2 College of Horticulture, Jilin Agricultural University, Changchun 130118, China; qiaojianlei918@163.com
* Correspondence: chenlimei@jlau.edu.cn; Tel.: +86-130-3901-0328
Abstract: In order to rapidly and accurately monitor cadmium contamination in leuce and under-
stand the growth conditions of leuce under cadmium pollution, leuce is used as the test material.
Under dierent concentrations of cadmium stress and at dierent growth stages, relative chloro-
phyll content of leuce leaves, the cadmium content in the leaves, and the visible-near infrared re-
ectance spectra are detected and analyzed. An inversion model of the cadmium content and rela-
tive chlorophyll content in the leuce leaves is established. The results indicate that cadmium con-
centrations of 1 mg/kg and 5 mg/kg promote relative chlorophyll content, while concentrations of
10 mg/kg and 20 mg/kg inhibit relative chlorophyll content. The cadmium content in the leaves
increases with increasing cadmium concentrations. Cadmium stress caused a “blue shift” in the red
edge position only during the mature period, while the red valley position underwent a “blue shift”
during the seedling and growth periods and a “red shift” during the mature period. The green peak
position exhibited a “blue shift”. After model validation, it was found that the model constructed
using the ratio of red edge area to yellow edge area and the normalized values of red edge area and
yellow edge area eectively estimated the cadmium content in leuce leaves. The model established
using the normalized vegetation index of the red edge and the ratio of the peak green value to red
shoulder amplitude can eectively estimate the relative chlorophyll content in leuce leaves. This
study demonstrates that the visible-near infrared spectroscopy technique holds great potential for
monitoring cadmium contamination and estimating chlorophyll content in leuce.
Keywords: spectral parameters; cadmium; inversion model; regression analysis; SPAD value
1. Introduction
In China, approximately 20% of cultivated land is aected by heavy metal pollution
[1]. Among this land, the rate of excessive cadmium pollution is 7%, making it the primary
pollutant in contaminated soil areas [2]. Cadmium is characterized by its high toxicity,
strong mobility, and resistance to degradation. It is easily absorbed and accumulated by
plants, and it can enter the human body through the food chain, posing a threat to human
health [3]. Therefore, rapid and accurate monitoring or identication of heavy metal stress
levels in plants is of signicant importance for ensuring food safety.
Currently, monitoring of plant heavy metal pollution is predominantly carried out
using chemical and spectroscopic techniques. Among them, chemical methods oer high
accuracy but are complex, destructive, and not suitable for rapid and large-scale monitor-
ing, requiring signicant human and material resources [4]. Spectral techniques can detect
Citation: Zhou, L.; Zhou, L.; Wu, H.;
Kong, L.; Li, J.; Qiao, J.; Chen, L.
Analysis of Cadmium
Contamination in Leuce (Lactuca
sativa L.) Using Visible-Near Infrared
Reectance Spectroscopy.
Sensors 2023, 23, 9562.
hps://doi.org/10.3390/
s23239562
Academic Editor: Simone Borri
Received: 10 November 2023
Revised: 29 November 2023
Accepted: 30 November 2023
Published: 1 December 2023
Copyright: © 2023 by the authors. Li-
censee MDPI, Basel, Swierland.
This article is an open access article
distributed under the terms and con-
ditions of the Creative Commons At-
tribution (CC BY) license (hps://cre-
ativecommons.org/licenses/by/4.0/).
Sensors 2023, 23, 9562 2 of 14
the levels of heavy metal pollution in plants and assess their growth conditions by ana-
lyzing changes in the reectance spectra of plant leaves. This allows for rapid and non-
destructive monitoring of heavy metal contamination in plants [5].
Many researchers have conducted in-depth studies on the monitoring of heavy metal
pollution in plants and the inversion of chlorophyll content using the variation character-
istics of plant spectra. Li et al. [6] analyzed the hyperspectral response characteristics of
chicory leaves under cadmium stress and established a monitoring model for the cad-
mium mass ratio in chicory leaves using rst-order derivative spectroscopy–partial least
squares regression (FDR-PLS). They believed that in situ hyperspectral technology could
achieve rapid and accurate monitoring of cadmium pollution. Abdel-Rahman et al. [7]
used hyperspectral data and partial least squares (PLS) to establish a quantitative moni-
toring model for cadmium content in Swiss chard leaves, achieving good prediction re-
sults. Chen et al. [8] analyzed the hyperspectral characteristics of tobacco leaves under
cadmium stress. The results showed that the ratio vegetation index (RVI) and normalized
dierence vegetation index (NDVI) in the spectral indices had a signicant correlation
with cadmium content in tobacco leaves and had a good predictive eect on the cadmium
content in tobacco leaves. Liu et al. [9] established a model for the inversion of chlorophyll
content in soybean leaves using spectral indices. The spectral indices dierence index (DI)
and rst-derivative dierence index (FDDI) showed the highest correlation with chloro-
phyll content, and the resulting model can provide a reference for large-scale monitoring
of soybean growth conditions. Wang et al. [10] used spectral characteristics to invert the
chlorophyll content in Sabina vulgaris leaves. The vegetation indices (NDVI) and vegeta-
tion cover index (mNDVI) showed a high correlation with the chlorophyll content, and
the inversion model established had a high accuracy.
In order to achieve rapid and accurate monitoring of cadmium pollution in leuce
and understand the growth status of leuce under cadmium contamination, this study
conducted detection and analysis of visible-near infrared reectance spectra, cadmium
content, and SPAD values in leuce leaves under cadmium pollution. A model for the
inversion of cadmium content and chlorophyll content in leuce leaves under cadmium
pollution was constructed. The purpose is to provide a theoretical basis and reference for
the safe production and quality control of leuce.
2. Materials and Methods
2.1. Materials and Experimental Design
The experiments were conducted in May 2023 within the campus of Jilin Agricultural
University in Changchun, Jilin Province (125°42′ E, 43°82′ N). The pot planting method
was employed, using leuce (Lactuca sativa L. cv. Grand Rapids) as the test material, pur-
chased from Kuishou Agricultural Technology Co., Ltd. (Langfang, China). The experi-
mental soil used was uncontaminated nutrient soil with an organic maer content of 12.59
g/kg, total nitrogen content of 0.727 g/kg, available phosphorus content of 0.007 g/kg, and
available potassium content of 0.15 g/kg. The soil was sieved with a screen to remove im-
purities and separate it into ne particles. Then, the soil was allowed to stand in a dry,
ventilated area for 3 days. The cadmium content in the experimental soil was set at 5 con-
centration gradients: 0 (control group, CK), 1, 5, 10, and 20 mg/kg. Distilled water was
used as the solvent, and cadmium nitrate was used as the external source of cadmium for
addition. A 200 mL solution of each concentration of cadmium was prepared. The dier-
ent concentrations of cadmium solution were sprayed layer-by-layer onto the correspond-
ing experimental soil and thoroughly mixed by turning over the soil. After aging for 10
days, the soil was transferred into pots with dimensions of 480 cm × 230 cm × 160 cm, with
1.5 kg of soil being added to each pot [11,12]. Three replicates were set for each treatment
level. When the leuce seedlings reached the stage of two leaves and one heart, uniformly
growing and healthy seedlings were selected and transplanted into the pots. Three seed-
lings were planted in each pot. To ensure the normal growth of the leuce, an adequate
Sensors 2023, 23, 9562 3 of 14
water supply was provided throughout the experiment. The positions of the pots were
changed every two days throughout the entire experimental period to ensure even expo-
sure to light.
2.2. Spectral Data Acquisition
The visible-near infrared reectance spectra data of leuce leaves were measured at
three dierent time points: 15 days (seedling stage) after cadmium stress, 30 days (growth
stage) after cadmium stress, and 45 days (maturity stage) after cadmium stress. The meas-
urements were conducted for leaves from the ve treatment groups. The spectrometer
selected for this study was the AvaSpec-ULS2048 multi-purpose ber optic spectrometer,
manufactured by Aventes in the Netherlands. This spectrometer has a wavelength range
of 200–1100 nm and a spectral resolution of 0.05–20 nm. The light source used was the
AvaLight-DHc full-range compact light source, also produced by Aventes. The deuterium
lamp covers the wavelength range of 200–400 nm, while the tungsten halogen lamp covers
the wavelength range of 400–2500 nm.
The measurements were conducted under the following weather conditions: clear
sky, no wind, and few clouds. The measurements were taken between 10:00 AM and 2:00
PM. During data collection, the rst step was to connect the ber optic probes separately
to the spectrometer and light source. Then, the reection probe was secured using a re-
ection probe holder in such a way that the angle between the reection probe and the
leaf surface was 45°. Finally, employing a deuterium–halogen lamp as the light source that
was allowed to preheat for 8 min, white balance calibration and measurements were per-
formed. White balance calibration was conducted every 30 min throughout the entire ex-
perimental process. For sample selection, the top 2 or top 3 leaves were chosen for meas-
urement. During the measurement, the main veins of the leaves were avoided. Eight spec-
tra data were collected for each leuce leaf, and any abnormal spectra data were removed.
Finally, the average value of the remaining spectra data was taken as the reection spec-
trum data for that particular leuce leaf. Under each cadmium stress treatment at dierent
growth stages, three spectral curves were obtained, resulting in a total of 45 curves.
2.3. SPAD Value Determination
The SPAD values of the leaves were determined using a handheld chlorophyll meter
(SPAD-502Plus, konica minolta,, Chiyodaku, Japan). The top 2 or top 3 leaves were se-
lected for measurement while avoiding the main leaf veins. Each leaf was measured three
times, and the average of these measurements was taken as the SPAD value for that par-
ticular leaf.
2.4. Leaf Cadmium Content Determination
The measurement of leaf cadmium content was conducted using a digestion method.
Firstly, the corresponding leaves for spectral data collection were washed, dried, and
ashed. Then, a Touchwin 2.0 microwave digestion instrument was used for digestion. The
specic steps were as follows: weighing 0.15 g of the sample into a 30 mL polytetrauoro-
ethylene crucible, adding 8 mL of a mixed acid solution (concentrated nitric acid:hydro-
uoric acid:perchloric acid = 3:3:1), heating with an open lid at 250 °C until the perchloric
acid volatilized (the sample was completely dried at this point), and then turning o the
power. When the temperature dropped to approximately 180 °C, 8 mL of aqua regia and
10 mL of internal standard mixed solution were added to the crucible, mixed well, ltered,
and set aside. Finally, 250 µL of the digestion solution was taken, mixed with 5 mL of 3%
nitric acid solution, and measured for cadmium content in the leaves using an inductively
coupled plasma–mass spectrometry (ICP-MS) instrument (300D, Perkinelmer, Waltham,
MA, USA).
Sensors 2023, 23, 9562 4 of 14
2.5. Data Processing and Analysis
2.5.1. Data Processing Methods
The AvaSoft 8.5 spectral acquisition software was used for collecting spectral data.
The collected spectral data were preprocessed and subjected to correlation analysis using
Origin 2021 software. SPSS 23 software was employed for conducting signicance anal-
yses and regression analyses on the data.
The Pearson correlation coefficient method was employed to assess the correlation be-
tween the leaf cadmium content, relative chlorophyll content (SPAD value), and spectral
parameters. Strongly correlated parameters were identified, with R-values closer to 1 or −1
indicating stronger correlations, while values closer to 0 indicated weaker correlations.
To determine if there are signicant dierences in the mean SPAD values among dif-
ferent groups, a signicance test is used. Typically, the signicance levels of 5% and 1%
are employed as criteria. If the signicance level is less than 5% or 1%, the dierence is
considered to be signicant or highly signicant, respectively.
The regression model can be utilized in various elds such as prediction, association
analysis, and causal inference. Therefore, in this study, we chose to employ a regression
model to investigate the relationship between the leaf cadmium content, SPAD value, and
spectral characteristic parameters. The goodness of t of the model for inferring leaf cad-
mium content from SPAD values is assessed based on the F-value and R-squared (R2). A
larger F-value indicates a beer t of the model, reecting a stronger explanatory power
of the model for the dependent variable. The closer R2 is to 1, the beer the regression
model ts the observed values, indicating a closer relationship between the two variables.
2.5.2. Calculation Method for Spectral Characteristic Parameters
In this study, spectral characteristic parameters such as the green peak, red valley,
red edge, red edge amplitude, and normalized dierence vegetation index at 705 nm
(NDVI705) were used for analysis. The specic calculation methods are shown in Table 1.
Table 1. Calculation methods for spectral characteristic parameters.
Feature Parameter
Parameter Description
Wavelength/nm
Pg (Peak green value)
Maximum value of leaf reflectance spectrum
500~600
Vr (Depth of red valley)
Minimum value of leaf reflectance spectrum
600~720
Pr (Red shoulder amplitude)
Maximum value of leaf reflectance spectrum
750~950
Λr (Red edge position)
Wavelength corresponding to the maximum value of the first-order de-
rivative of the leaf reflectance spectrum
670~780
Dr (Amplitude of red edge)
Maximum value of the first-order derivative of the leaf reflectance spec-
trum
670~780
Drmin (Minimum amplitude of red
edge)
Minimum value of the first-order derivative of the leaf reflectance spec-
trum
670~780
Dy (Amplitude of yellow edge)
Maximum value of the first-order derivative of the leaf reflectance spec-
trum
560~640
Db (Amplitude of blue edge)
Maximum value of the first-order derivative of the leaf reflectance spec-
trum
490~530
SDr (Red edge area)
Sum of the first-order derivative of the leaf reflectance spectrum
670~780
SDy (Yellow edge area)
Sum of the first-order derivative of the leaf reflectance spectrum
560~640
SDb (Blue edge area)
Sum of the first-order derivative of the leaf reflectance spectrum
490~530
SDr/SDy
Red edge area/yellow edge area
/
SDr/SDb
Red edge area/blue edge area
/
SDb/SDy
Blue edge area/yellow edge area
/
(SDr − SDy)/(SDr + SDy)
Normalized value of red edge area and yellow edge area
/
(SDr − SDb)/(SDr + SDb)
Normalized value of red edge area and blue edge area
/
(SDb − SDy)/(SDb + SDy)
Normalized value of blue edge area and yellow edge area
/
NDVI705 (Normalized vegetation index
of the red edge)
）（）（ 705750705750705 /RRRRNDVI +−=
R750 and R705 repre-
sent the spectral
Sensors 2023, 23, 9562 5 of 14
reflectance values
at 750 nm and 705
nm, respectively
Dr/Drmin
Amplitude of red edge/minimum amplitude of red edge
/
Pg/Pr
Peak green value/red shoulder amplitude
/
Dr/Dy
Amplitude of red edge/amplitude of yellow edge
/
Dr/Db
Amplitude of red edge/amplitude of blue edge
/
Dy/Db
Amplitude of yellow edge/amplitude of blue edge
/
3. Results and Discussion
3.1. Eects of Cadmium Stress on the Biochemical Parameters of Leuce
3.1.1. Eects of Cadmium Stress on the SPAD Value of Leuce
Figure 1 shows the changes in SPAD values in leuce leaves at dierent growth
stages under cadmium stress. During the seedling stage, the SPAD values increased with
increasing cadmium concentrations in the soil. During the growth stage, the SPAD values
increased in the 1 mg/kg and 5 mg/kg cadmium stress groups compared to the control
group (CK) but decreased in the 10 mg/kg and 20 mg/kg cadmium stress groups com-
pared to the control group. During the mature stage, the SPAD values of the leuce leaves
gradually decreased with increasing cadmium concentrations in the soil. This study used
cadmium nitrate as an exogenous source of cadmium. It is worth noting that nitrate ions
can promote the accumulation of chlorophyll in leuce leaves [13]. However, it should
also be noted that nitrate ion and cadmium ion concentrations vary in direct proportion,
and under the stress of 10 mg/kg and 20 mg/kg cadmium nitrate, there is an inhibitory
eect on the SPAD values. Moreover, similar to the research results obtained by Jia et al.
[14–16], who used cadmium chloride as an exogenous source of cadmium, the stress of 1
mg/kg and 5 mg/kg cadmium nitrate also exhibited a promoting eect on the SPAD val-
ues. Therefore, it can be inferred that under cadmium nitrate stress, cadmium plays a
dominant role in aecting the SPAD values. The stress of 1 mg/kg and 5 mg/kg cadmium
promotes the SPAD values, possibly due to the ability of metal ions to enhance the metab-
olism of cytokinin enzymes, thereby promoting cell growth and increasing chlorophyll
content. On the other hand, the stress of 10 mg/kg and 20 mg/kg cadmium inhibits the
SPAD values, likely because cadmium stress can cause damage to the chloroplast struc-
ture in plant leaf cells, leading to hindered biosynthesis of chlorophyll [15–17].
Figure 1. Eects of cadmium stress on the SPAD values of leuce. Note: Dierent lowercase leers
indicate signicant dierences at the p < 0.05 level.
Sensors 2023, 23, 9562 6 of 14
3.1.2. Eects of Cadmium Stress on Cadmium Content in Leuce Leaves
According to Figure 2, the cadmium content in the leuce leaves at dierent growth
stages showed an increasing trend with the increasing cadmium concentrations in the soil.
Leuce is sensitive to cadmium stress and can eectively accumulate cadmium. The en-
richment eect becomes more pronounced with increasing cadmium concentrations. This
is aributed to the presence of substances such as organic acids, cellulose, and pectin in
the plant cell wall, which can chelate cadmium. As the cadmium concentration increases,
the content of cadmium in the cell organelles increases and accumulates continuously
[18,19].
Figure 2. Eects of cadmium stress on the cadmium content in leuce leaves.
3.2. Analysis of Spectral Response Characteristics of Leuce Leaves under Cadmium Stress
3.2.1. Dierential Analysis of Visible-Near Infrared Spectra
Vegetation is sensitive to stress in the visible-near infrared band, and spectral bands
beyond 1000 nm are greatly aected by water vapor absorption. Therefore, this paper
takes the spectral range of 400–980 nm as the analysis object [20,21]. The utilization of the
SG smoothing technique facilitates a more continuous and gradual transition between
data points while still retaining the trend and features of the original curve. This approach
proves advantageous for subsequent analysis [22]. Therefore, the raw spectral data col-
lected from the ve treatments were subjected to SG smoothing, with the parameters set
to a second-degree polynomial and 11 smoothing points [23], resulting in the visible-near
infrared reectance spectra curves of the leuce leaves at dierent growth stages under
cadmium stress shown in Figure 3.
The visible-near infrared reectance spectra of the leuce leaves under dierent cad-
mium concentrations and growth stages exhibit a high degree of similarity in their char-
acteristic changes. At approximately 670 nm, a distinct absorption valley, known as the
“red valley,” is observed, while a reection peak, known as the “green peak,” appears
close to 550 nm. In the wavelength range of 680 nm–750 nm, the reectance of the leuce
leaves increases sharply, showing a typical “red edge eect”. During the growth and mat-
uration stages, a small absorption valley is observed at approximately 760 nm in the let-
tuce leaves, which could be aributed to the narrow water absorption band in that wave-
length range [24,25].
Sensors 2023, 23, 9562 7 of 14
(a)
(b)
(c)
Figure 3. Spectral characteristics of reectance in leuce leaves under cadmium stress: (a) spectral
characteristics of reectance in leuce leaf seedlings under cadmium stress; (b) spectral characteris-
tics of reectance in leuce leaves during the growth stage under cadmium stress; (c) spectral char-
acteristics of reectance in leuce leaves during the maturation stage under cadmium stress.
3.2.2. Analysis of Spectral Characteristic Parameters in Leuce Leaves
Based on the analysis of the dierences in the visible-near infrared reectance spec-
tra, representative spectral features such as the red edge, red valley, and green peak were
selected for further analysis. As shown in Table 2, compared to the CK group, the red edge
position showed no change during the seedling stage and growth stage but exhibited a 9
nm “blue shift” during the mature stage. The red valley position initially experienced a
“blue shift” followed by a “red shift” (6 nm blue shift during the seedling stage, 7 nm blue
shift during the growth stage, and 5 nm red shift during the mature stage). The green peak
position underwent a “blue shift” (3 nm blue shift during the seedling stage, 1 nm blue
shift during the growth stage, and 3 nm blue shift during the mature stage). When plants
are subjected to heavy metal stress, the activity of enzymes required for chlorophyll for-
mation within the plant is inhibited, hindering chlorophyll formation. This leads to an
increase in lutein and a decrease in chlorophyll, resulting in a “blue shift” in the spectral
features of the red edge and red valley [26,27]. The “red shift” in the red valley position
and the “blue shift” in the green peak position during the maturation stage may be related
to self-regulation or changes in the leaf structure of leuce plants under cadmium stress
[28,29]. Therefore, the red edge, red valley, and green peak can be used to discriminate
the extent of cadmium pollution in leuce plants.
Sensors 2023, 23, 9562 8 of 14
Table 2. Spectral characteristic parameters of leuce leaves at dierent growth stages under dier-
ent cadmium concentrations.
Cadmium
Treatment
Seedling Stage
Growth Stage
Mature Stage
Red Edge
Position/nm
Green Peak
Position/nm
Red Valley
Posi-
tion/nm
Red Edge
Position/nm
Green Peak
Position/nm
Red Valley
Position/nm
Red Edge
Position/nm
Green Peak
Posi-
tion/nm
Red Valley
Position/nm
CK
702
551
676
693
548
677
702
550
671
Cd1
702
551
676
693
548
678
702
550
671
Cd5
702
548
673
693
548
676
702
551
671
Cd10
702
551
670
693
548
670
693
551
676
Cd20
702
551
670
693
547
676
702
547
676
3.2.3. Analysis of Normalized Dierence Vegetation Index (NDVI705) at the Red Edge
The normalized dierence vegetation index (NDVI705) at the red edge is highly sensi-
tive to the environment and is one of the commonly used indicators for detecting plant
stress. Its value ranges from −1 to 1, with a typical range of 0.2 to 0.9 for green vegetation
areas. It is generally considered that when the NDVI705 value is less than 0.2, it indicates
that the plant is under stress [27,30]. In this study, the NDVI705 decreased rst and then
increased with increasing cadmium concentrations during the seedling stage, but its value
was lower than that of the CK group. During the growth stage, the NDVI705 rst increased
and then decreased with increasing of cadmium concentrations. During the maturation
stage, the NDVI705 decreased with increasing cadmium concentrations (Figure 4). Under
cadmium stress conditions of 10 mg/kg and 20 mg/kg, the NDVI705 values at the dierent
growth stages were lower than those of the CK group, indicating that the cadmium stress
of 10 mg/kg and 20 mg/kg had an impact on the NDVI705 values of the leuce and exerted
a stress eect on the growth of the leuce [31].
Figure 4. Normalized dierence vegetation index (NDVI705) of leuce under dierent cadmium con-
centrations at dierent growth stages.
3.3. Correlation Analysis between Spectral Characteristic Parameters, Leaf Cadmium Content,
and SPAD Value
To screen for sensitive spectral characteristic parameters of leuce under cadmium
stress, a correlation analysis was conducted between 23 spectral characteristic parameters,
the leaf cadmium content, and the SPAD value. The results are depicted in Figure 5, where
red indicates a positive correlation between two parameters and blue represents a nega-
tive correlation. The size of the circles represents the degree of association, with larger
circles indicating higher correlations. Among the 23 characteristic parameters, SDy, SDb,
SDr/SDy, SDb/SDy, (SDr − SDy)/(SDr + SDy), (SDb − SDy)/(SDb + SDy), and Dr/Dy
showed high correlations with leaf cadmium content. The research ndings indicate that
Sensors 2023, 23, 9562 9 of 14
the spectral characteristic parameters of the “three-edge” region in leuce leaves are sen-
sitive to cadmium stress. The accumulation of cadmium in plant leaves can be reected in
their physiology through various physiological indicators, such as chlorophyll and cell
structure [17]. Slight changes in these indicators can cause alterations in the spectral char-
acteristics of the leaves. In addition, the spectral characteristics of the red edge and yellow
edge regions can also serve as diagnostic indicators for cadmium pollution in leuce [32].
The spectral characteristic parameters of the “three-edge” region are eective wave-
lengths for monitoring the extent of heavy metal Pb2+ pollution, while the blue edge and
red edge are eective wavelengths for monitoring the extent of heavy metal Cu2+ pollution
[33]. Therefore, the spectral characteristic parameters associated with chlorophyll and cell
structure can serve as indicators for diagnosing cadmium pollution in leuce.
Pg, NDVI705, and Pg/Pr showed high correlations with the SPAD value. Using ratio
and normalization methods to construct spectral characteristic parameters has shown bet-
ter results in estimating leaf chlorophyll content [34]. NDVI is eective in reecting crop
growth and eliminating partial radiation errors, making it suitable for dynamic vegetation
monitoring [35]. Pg can serve as an important spectral parameter for estimating the SPAD
value of tomato leaves under disease stress [36]. The screened spectral characteristic pa-
rameters in this study all reached a signicant level of correlation with the SPAD value,
demonstrating statistical signicance. They can be regarded as important spectral charac-
teristic parameters for accurately estimating the chlorophyll content in leuce leaves un-
der cadmium stress.
Figure 5. Correlation analysis between spectral characteristic parameters, leaf cadmium content,
and SPAD value under cadmium stress.
3.4. Establishment of a Leaf Cadmium Content Inversion Model for Lettuce under Cadmium Stress
According to the results of the correlation analysis, using 45 collected spectral data
points as the modeling samples, a regression analysis was conducted using leuce leaf
cadmium content as the dependent variable and spectral characteristic parameters SDy,
Sensors 2023, 23, 9562 10 of 14
SDb, SDr/Sdy, SDb/Sdy, (SDr − Sdy)/(SDr + Sdy), (SDb − Sdy)/(SDb + Sdy), Dr/Dy as inde-
pendent variables. Linear, quadratic, cubic, and logarithmic functions were ed to the
model. The results are presented in Table 3. The cadmium content in the leuce leaves
showed a high correlation with SDr/SDy, SDb/SDy, and (SDr − SDy)/(SDr + SDy), with
ing coecients (R2) of 0.872, 0.781, and 0.792, respectively.
Table 3. Inversion models for cadmium content in lettuce leaves under cadmium contamination stress.
Spectral Characteristic Pa-
rameters
Fitting Model
R2
Significance
F
SDy
32 001.0033.0053.0305.12yxxx +−+=
0.535
*
4.223
SDb
2
011.0286.0265.1y xx ++=
0.431
*
4.545
SDr/SDy
32 009.0305.0256.3934.3y xxx +−+−=
0.872
**
24.959
SDb/SDy
32 973.0462.8988.19007.2y xxx +−+−=
0.781
**
13.041
(SDr − SDy)/(SDr + SDy)
32 942.26829.6599.0186.0y xxx +−−=
0.792
**
13.996
(SDb − SDy)/(SDb + SDy)
32 209.14488.7075.15545.10yxxx −−+=
0.65
**
6.800
Dr/Dy
2
561.13732.47046.32yxx −+−=
0.463
*
5.178
Note: * and ** indicate signicant correlations at the 0.05 and 0.01 levels, respectively.
SDr/SDy, SDb/SDy, and (SDr − SDy)/(SDr + SDy) demonstrated beer performance
in estimating cadmium content in leuce leaves. However, a study by Gu et al. [32] sug-
gests that the red edge area SDr has the best performance in estimating cadmium content
in Chinese cabbage leaves. Chen et al. [8] found that the normalized vegetation index
NDVI is eective in estimating cadmium content in tobacco leaves. Zhong et al. [37] con-
cluded that NDVI has the best performance in estimating cadmium content in various
organs of rice. The reason for this dierence may lie in the variation of the tested materials.
It suggests that dierent plant species may have dierent spectral characteristic parame-
ters that are sensitive to cadmium.
3.5. Model for SPAD Value Inversion in Leuce Leaves under Cadmium Stress
The selected spectral parameters, including Pg, NDVI705, and Pg/Pr, were combined
with the leuce SPAD values for analysis. Linear, quadratic, and cubic regression models
were used to t the SPAD value models. The results are presented in Table 4, where the
ing coecients (R2) between the leuce leaf SPAD values and Pg, NDVI705, and Pg/Pr
were 0.677, 0.789, and 0.755, respectively.
Table 4. Inversion models for SPAD values in leuce leaves under cadmium contamination.
Spectral Charac-
teristic Parameters
Fitting Model
R2
Significance
F
Pg
2
036.0-043.1548.21 xxy +=
0.677
**
12.569
NDVI705
2
359.259046.94196.29yxx +−=
0.789
**
22.407
Pg/Pr
x575.56787.47y−=
0.755
**
40.009
Note: ** indicate signicant correlation at the 0.01 levels.
By constructing spectral indices, redundant spectral information can be eectively
removed, thereby improving the accuracy of estimating plant chlorophyll content [38]. In
this study, the model for estimating SPAD values using NDVI705 as the independent vari-
able showed the best ing eect. This is consistent with the ndings of Guo et al. [39]
regarding the estimation of corn SPAD values using hyperspectral data, where the original
spectral NDVI was identied as the optimal parameter for estimating corn SPAD values.
Sensors 2023, 23, 9562 11 of 14
However, this is in contrast to the ndings of He et al. [34], who reported that the normal-
ized vegetation index had an R2 value of 0 when used for quantitative estimation of chlo-
rophyll content in leaves of karst plants. This dierence may arise from the variations in
environmental conditions and plant species. Therefore, it is necessary to select spectral
parameters for estimating SPAD values based on the environmental conditions and plant
species.
3.6. Model Validation
The spectral data for ve samples each in the seedling stage, growth stage, and ma-
ture stage were randomly selected to calculate SDr/SDy, SDb/SDy, (SDr − SDy)/(SDr +
SDy), Pg, NDVI705, and Pg/Pr as six spectral characteristic parameters. SDr/SDy, SDb/SDy,
and (SDr − SDy)/(SDr + SDy) were used as independent variables to predict the cadmium
content in the leuce leaves, while Pg, NDVI705, and Pg/Pr were used as independent var-
iables to predict the SPAD value of the leuce leaves. The predicted values were compared
with the actual values, and the results are shown in Figure 6. The SDr/SDy and (SDr −
Sdy)/(SDr + Sdy) models can accurately estimate cadmium content in leuce leaves. The
NDVI705 and Pg/Pr models can accurately estimate the SPAD values of leuce leaves.
Figure 6. Fied curve of predicted values and actual measurements.
Sensors 2023, 23, 9562 12 of 14
4. Conclusions
Leuce was used as the research object, and a pot planting method with exogenous
addition of cadmium was used to study the leuce leaf SPAD values, cadmium content in
the leaves, and visible-near-infrared reectance spectra under cadmium stress. The fol-
lowing conclusions were drawn:
(1) Under cadmium stress, the SPAD values of the leuce leaves exhibited inhibition at
high concentrations and promotion at low concentrations. Moreover, the cadmium
concentration in the leuce leaves increased with increasing soil cadmium concen-
trations.
(2) The red edge, red valley, and green peak positions in the reectance spectra of the
leuce leaves were sensitive to cadmium stress. Under cadmium stress, these posi-
tions shifted, and they could be used for preliminary diagnosis of cadmium pollution
in leuce. The normalized dierence vegetation index of the red edge (NDVI705) was
lower than the CK group under 10 mg/kg and 20 mg/kg cadmium stress, indicating
that leuce growth was aected by the cadmium stress.
(3) The models corresponding to SDr/SDy and (SDr − SDy)/(SDr + SDy) can eectively
estimate the cadmium content in leuce leaves. The models corresponding to NDVI705
and Pg/Pr can accurately estimate the SPAD values of leuce leaves.
This research demonstrates that the utilization of visible-near infrared spectroscopy
technology can provide a theoretical basis and reference for the safe production and qual-
ity control of leuce.
Author Contributions: Conceptualization, L.Z. (Lina Zhou) and L.Z. (Leijinyu Zhou); methodology,
L.Z. (Leijinyu Zhou) and L.C.; formal analysis, L.Z. (Leijinyu Zhou) and L.K.; investigation, L.C. and
J.Q.; resources, L.Z. (Lina Zhou) and J.L.; writing—original draft preparation, L.Z. (Leijinyu Zhou) and
H.W.; writing—review and editing, J.L.; visualization, L.K. and J.Q.; supervision, L.C.; funding acqui-
sition, L.Z. (Lina Zhou). All authors have read and agreed to the published version of the manuscript.
Funding: This work was supported by the Science and Technology Development Project of Jilin
Province, grant number 20210202051NC, founded by Lina Zhou. This work was also supported by
the Science and Technology Development Project of Jilin Province, grant number 20200403140SF,
founded by Jianlei Qiao.
Institutional Review Board Statement: This study did not require ethical approval.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data that support the ndings of this study are available from the
authors upon reasonable request.
Conicts of Interest: The authors declare no conict of interest.
References
1. Pandey, C.V.; Bajpai, O.; Singh, N. Energy crops in sustainable phytoremediation. Renew. Sustain. Energy Rev. 2016, 54, 58–73.
2. Zhao, F.; Ma, Y.; Zhu, Y.; Tang, Z.; McGrath, S.P. Soil Contamination in China: Current Status and Mitigation Strategies. Environ.
Sci. Technol. 2014, 49, 750–759.
3. Liang, X.N.; Liang, S.; Liang, Y.P.; Zhang, J. Eects of exogenous Ca on seed germination and growth of Miscanthus sacchariorus
seedlings under Cd stress. Chin. J. Ecol. 2023, 1–10.
4. Zhong, L.; Qian, J.W.; Chu, X.Y.; Qian, Z.H.; Wang, M.; Li, J.L. Monitoring heavy metal pollution contamination of wheat soil
using hyperspectral remote sensing technology. Trans. Chin. Soc. Agric. Eng. 2023, 39, 265–270.
5. Liu, Y.P.; Luo, Q.; Cheng, H.F. Application and development of hyperspectral remote sensing technology to determine the
heavy metal content in soil. J. Agro-Environ. Sci. 2020, 39, 2699–2709.
6. Li, L.T.; Shen, F.M.; Ma, W.L.; Fan, J.; Li, Y.R.; Liu, H.T. Response Characteristics and Quantitative Monitoring Models Analyzed
Using in situ Leaf Hyperspectra under Dierent Cd Stress Conditions. Trans. Chin. Soc. Agric. Mach. 2020, 51, 146–155.
7. Abdel-Rahman, M.E.; Mutanga, O.; Odindi, J.; Adam, E.; Odindo, A.; Ismail, R. Estimating Swiss chard foliar macro- and mi-
cronutrient concentrations under dierent irrigation water sources using ground-based hyperspectral data and four partial least
squares (PLS)-based (PLS1, PLS2, SPLS1 and SPLS2) regression algorithms. Comput. Electron. Agric. 2016, 132, 21–33.
8. Chen, N.; Feng, H.L.; Yang, Y.D.; Chen, P.; Ren, T.B.; Jia, F.F.; Liu, G.S. Establishment of hyperspectral prediction model for
cadmium content in ue-cured tobacco leaves. J. Agric. Resour. Environ. 2021, 38, 570–575.
Sensors 2023, 23, 9562 13 of 14
9. Liu, S.; Yu, H.Y.; Zhang, J.H.; Zhou, H.G.; Kong, L.J.; Zhang, L.; Dang, J.M.; Sui, Y.Y. Study on Inversion Model of Chlorophyll
content in Soybean leaf Based on Optimal Spectral Indices. Spectrosc. Spectr. Anal. 2021, 41, 1912–1919.
10. Wang, N.; Yang, G.; Han, X.; Jia, G.; Li, Q.; Liu, F.; Liu, X.; Chen, H.; Guo, X.; Zhang, T. Study of the spectral characters–chloro-
phyll inversion model of Sabina vulgaris in the Mu Us Sandy Land. Front. Earth Sci. 2023, 10, 1032585.
11. Tao, L.; Zhang, N.M. Growth Response of Three Leafy Vegetables to Cd Pollution and Their Cd Accumulation Characteristics.
Chin. Agric. Sci. Bull. 2018, 34, 99–106.
12. Sun, J.; Zhang, Y.C.; Mao, H.P.; Wu, X.H.; Chen, Y.; Wen, Q.P. Responses Analysis of Leuce Leaf Pollution in Cadmium Stress
Based on Computer Vision. Trans. Chin. Soc. Agric. Mach. 2018, 49, 166–172.
13. Coskun, D.; Brio, D.T.; Shi, W.; Kronzucker, H.J. How Plant Root Exudates Shape the Nitrogen Cycle. Trends Plant Sci. 2017,
22, 661–673.
14. Jia, Y.H.; Han, Y.Y.; Liu, J.; Gao, F.; Liang, Q.; Yu, P.; Liu, C.J.; Zhang, X.; Su, B.W. Physiological adaptations to cadmium stresses
and cadmium accumulation in leuce. J. Agro-Environ. Sci. 2018, 37, 1610–1618.
15. Yang, Y.; Li, Y.L.; Chen, W.P.; Wang, M.E.; Peng, C. Variation Characteristics of Vegetables Cadmium Uptake Factors and Its
Relations to Environmental Factors. Environ. Sci. 2017, 38, 399–404.
16. Chen, X.J.; Tao, H.F.; Wu, Y.Z.; Xu, X. Eects of Cadmium on metabolism of photosynthetic pigment and photosynthetic system
in Lactuca sativa L. revealed by physiological and proteomics analysis. Sci. Hortic. 2022, 305, 111371.
17. Xu, J.; Hu, B.H.; Ge, T.; Chen, Q. Eects of Cadmium Stress on Seed Germination and Physiological Characteristics of Leuce
Seedling. Hubei Agric. Sci. 2014, 53, 4892–4896.
18. Du, Y.P.; Li, H.J.; Yin, K.L.; Zhai, H. Cadmium accumulation, subcellular distribution, and chemical forms in Vitis vinifera cv.
Chardonnay grapevine. Chin. J. Appl. Ecol. 2012, 23, 1607–1612.
19. Yu, P.; Gao, F.; Liu, J.; Liang, Q.; Han, Y.Y.; Wang, J.X.; Jia, Y.H. Eect of Cd on Plant Growth and Its Tolerance Mechanism.
Chin. Agric. Sci. Bull. 2017, 33, 89–95.
20. Zhang, C.; Su, X.Y.; Xia, T.; Yang, K.M.; Feng, F.S. Monitoring the Degree of Pollution in Dierent Varieties of Maize Under
Copper and Lead stress. Spectrosc. Spectr. Anal. 2023, 43, 1268–1274.
21. Li, Y.; Yang, K.M.; Rong, K.P.; Zhang, C.; Gao, P.; Cheng, F. Spectral Characteristics and Identication Research of Corn under
Copper Stress. Spectrosc. Spectr. Anal. 2019, 39, 2823–2828.
22. Zhou, X.; Sun, J.; Tian, Y.; Chen, Q.; Wu, X.; Hang, Y. A deep learning based regression method on hyperspectral data for rapid
prediction of cadmium residue in leuce leaves. Chemom. Intell. Lab. Syst. 2020, 200, 103996–103996; prepublish.
23. Xu, S.L.; Gao, Y.; Hu, G.L.; Yu, X.Z.; Zhang, R. Rapid Determination of Total Sugar Content of Goji Berries (Lycium barbarum)
by Near Infrared Spectroscopy with Eective Wavenumber Selection. Food Sci. 2016, 37, 105–109.
24. Liu, W.; Yu, Q.; Niu, T.; Yang, L.Z.; Liu, H.J.; Yan, F. Study on the Relationship Between Element as in Soil of Agricultural Land
and Leaf Spectral Characteristics. Spectrosc. Spectr. Anal. 2021, 41, 2866–2871.
25. Feng, W.; Guo, T.C.; Xie, Y.X.; Wang, Y.H.; Zhu, Y.J.; Wang, C.Y. Spectrum Analytical Technique and Its Applications for the
Crop Growth Detection. Chin. Agric. Sci. Bull. 2009, 25, 182–188.
26. Wang, W.; Shen, R.P.; Ji, C.X. Study on Heavy Metal Cu based on Hyperspectral Remote Sensing. Remote Sens. Technol. Appl.
2011, 26, 348–354.
27. Wang, H.; Zeng, L.S.; Sun, Y.H.; Zhang, J.H.; Guo, Q.Z.; Sun, F.L.; Song, C.Y.; Chen, J.M. Wheat canopy spectral reectance
feature response to heavy metal copper and zinc stress. Trans. Chin. Soc. Agric. Eng. 2017, 33, 171–176.
28. Kong, L.J.; Yu, H.Y.; Chen, M.C.; Piao, Z.J.; Liu, S.; Dang, J.M.; Zhang, L.; Sui, Y.Y. Analyze on the Response Characteristics of
Leaf vegetables to Particle Maers Based on Hyperspectral. Spectrosc. Spectr. Anal. 2021, 41, 236–242.
29. Jia, J.Z.; Fei, L.; Xin, D. Using Red Edge Position Shift to Monitor Grassland Grazing Intensity in Inner Mongolia. J. Indian Soc.
Remote Sens. 2018, 46, 81–88.
30. Hussein, O.S.; Kovács, F.; Tobak, Z. Spatiotemporal Assessment of Vegetation Indices and Land Cover for Erbil City and Its
Surrounding Using Modis Imageries. J. Environ. Geogr. 2017, 10, 31–39.
31. Che, Y.F.; Yan, D.; Wang, Z.Q.; Xiao, C. Spectral Eect Characteristics of Leuce under Strontium Stress. Sci. Technol. Eng. 2023,
23, 3544–3551.
32. Gu, Y.W.; Li, S.; Gao, W.; Wei, H. Hyperspectral estimation of the cadmium content in leaves of Brassica rapa chinesis based on
the spectral parameters. Acta Ecol. Sin. 2015, 35, 4445–4453.
33. Zhang, C.; Yang, K.M.; Wang, M.; Gao, P.; Cheng, F.; Li, Y. LD-CR-SIDSCAtan Detection Model for the Weak Spectral Information
of Maize Leaves under Copper and Lead Stresses. Spectrosc. Spectr. Anal. 2019, 39, 2091–2099.
34. He, W.; Yu, L.; Yao, Y.F. Estimation of plant leaf chlorophyll content based on spectral index in karst areas. Guihaia 2022, 42,
914–926.
35. Luo, J.; Yang, Z.Q.; Yang, L.; Yuan, C.H.; Zhang, F.Y.; Li, Y.C.; Li, C.Y. Establishment of an Estimation Model for Chlorophyll
Content of Strawberry Leaves under High Temperature Conditions at Seedling Stage Based on Hyperspectral Parameters. Chin.
J. Agrometeorol. 2022, 43, 832–845.
36. Xiang, Q.; Yang, Z.Q.; Wu, L.; Zhang, J.J.; Wei, W. Hyperspectral Estimation Model for SPAD Value of Tomato Leaf under Virus
Disease Infection. Chin. J. Agrometeorol. 2023, 44, 707–720.
37. Zhong, X.C.; Dai, Q.G.; He, L.; Chen, J.D.; Sun, C.M.; Gao, H.; Zhang, H.C.; Zheng, C. Rice Canopy Spectral Characteristics and
Its Forecast Evaluation Under Cadmium Stress. J. Agro-Environ. Sci. 2012, 31, 448–454.
Sensors 2023, 23, 9562 14 of 14
38. Yang, H.B.; Yin, H.; Li, F.; Hu, Y.; Yu, K. Machine learning models fed with optimized spectral indices to advance crop nitrogen
monitoring. Field Crops Res. 2023, 293, 108844.
39. Guo, S.; Chang, Q.R.; Cui, X.T.; Zhang, Y.M.; Chen, Q.; Jiang, D.Y.; Luo, L.L. Hyperspectral estimation of maize SPAD value
based on spectrum transformation and SPA-SVR. J. Northeast Agric. Univ. 2021, 52, 79–88.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual au-
thor(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.
