sensors

Article
Design and Test of a Soil Profile Moisture Sensor
Based on Sensitive Soil Layers
Zhenran Gao ID
, Yan Zhu ID
, Cheng Liu, Hongzhou Qian, Weixing Cao * and Jun Ni *
National Engineering and Technology Center for Information Agriculture,
Jiangsu Key Laboratory for Information Agriculture, Key Laboratory for Crop System Analysis and Decision
Making, Ministry of Agriculture, Jiangsu Key Laboratory for Information Agriculture,
Jiangsu Collaborative Innovation Center for the Technology and Application of Internet of Things,
Nanjing Agricultural University, Nanjing 210095, China; gzrsunrise@163.com (Z.G.);
yanzhu@njau.edu.cn (Y.Z.); lliuchemg@163.com (C.L.); zddlth@163.com (H.Q.)
* Correspondence: caow@njau.edu.cn (W.C.); nijun@njau.edu.cn (J.N.);
Tel.: +86-25-8439-6672 (W.C.); +86-25-8439-6593 (J.N.)




Received: 2 April 2018; Accepted: 16 May 2018; Published: 21 May 2018


Abstract: To meet the demand of intelligent irrigation for accurate moisture sensing in the soil
vertical profile, a soil profile moisture sensor was designed based on the principle of high-frequency
capacitance. The sensor consists of five groups of sensing probes, a data processor, and some
accessory components. Low-resistivity copper rings were used as components of the sensing probes.
Composable simulation of the sensor’s sensing probes was carried out using a high-frequency
structure simulator. According to the effective radiation range of electric field intensity, width and
spacing of copper ring were set to 30 mm and 40 mm, respectively. A parallel resonance circuit
of voltage-controlled oscillator and high-frequency inductance-capacitance (LC) was designed for
signal frequency division and conditioning. A data processor was used to process moisture-related
frequency signals for soil profile moisture sensing. The sensor was able to detect real-time soil
moisture at the depths of 20, 30, and 50 cm and conduct online inversion of moisture in the soil
layer between 0–100 cm. According to the calibration results, the degree of fitting (R2 ) between the
sensor’s measuring frequency and the volumetric moisture content of soil sample was 0.99 and the
relative error of the sensor consistency test was 0–1.17%. Field tests in different loam soils showed
that measured soil moisture from our sensor reproduced the observed soil moisture dynamic well,
with an R2 of 0.96 and a root mean square error of 0.04. In a sensor accuracy test, the R2 between the
measured value of the proposed sensor and that of the Diviner2000 portable soil moisture monitoring
system was higher than 0.85, with a relative error smaller than 5%. The R2 between measured values
and inversed soil moisture values for other soil layers were consistently higher than 0.8. According
to calibration test and field test, this sensor, which features low cost, good operability, and high
integration, is qualified for precise agricultural irrigation with stable performance and high accuracy.

Keywords: soil profile; high-frequency capacitance; moisture sensor; field test

1. Introduction
As a key factor influencing crop growth and yield, soil moisture content is an important basis for
the management of agricultural production [1]. To realize intelligent farmland irrigation, it is important
to measure soil profile moisture content accurately in real-time, to understand spatial water utilization,
and to provide a basis for selection of time and threshold of optimum irrigation and development
of precision irrigation [2,3]. Currently, the burial depth of soil moisture sensing sensor for moisture
measurement in the soil vertical profile has not been systematically investigated. Most soil moisture

Sensors 2018, 18, 1648; doi:10.3390/s18051648 www.mdpi.com/journal/sensors

Sensors 2018, 18, 1648 2 of 23

sensors are needle-type with a typical needle length of 3–20 cm [4], which makes them only suitable
for topsoil moisture content measurement rather than deep soil moisture measurement [5,6]. For the
purpose of soil vertical profile moisture measurement, multiple soil moisture sensors are usually buried
at different soil depths. This method not only requires substantial investment of time and effort, but the
in-situ soil measurement environment is also vulnerable to disturbance during sensor arrangement,
making it difficult to ensure consistency between the performance of multiple sensors [7–10].
To solve the difficulties of soil profile moisture measurement with the needle-type sensor,
non-contact soil profile moisture measurement methods have been investigated both in China and
abroad [11–18]. Yuan et al. [19] designed an equally spaced three-depth soil profile moisture sensor
based on the impedance method. With a radial measurement radius of 3–4.1 cm and equal spacing of
20 cm, this sensor would result in information redundancy in practice. A tubular dielectric moisture
sensor measurement system similar to EnviroSCAN was designed by Zhou et al. [20]; however,
the system requires the use of a motor for horizontal migration of the single-point sensor at a single
soil depth, making it inapplicable to automatic measurement of soil moisture content at different
depths. Similar to Diviner2000 (Sentek, Australia), the manual single-point soil profile moisture sensor
developed by Fu et al. [21] requires manual lifting as it is not able to achieve automatic moisture
measurement in the vertical profile. Currently, the products with the widest applications in this
field are the EnviroSCAN soil moisture profiling system [22,23] and the Diviner2000 soil moisture
profiler [24]; however, both of these are expensive.
In general, current soil profile moisture measurement requires large numbers of soil moisture
sensors at different soil layers, whereas data redundancy, a cumbersome process, and high cost make
existing equal-spacing soil profile moisture sensors incapable. Given the lack of detailed information
relating to soil profile moisture, intelligent irrigation is in great need of a low-cost and high-efficiency
soil profile moisture sensing system to reduce the cost of soil moisture measurement and enhance
irrigation efficiency [25,26].
Non-destructive sensing systems for soil moisture include time domain reflectometry (TDR),
frequency domain reflectometry (FDR), and high-frequency capacitance (HFC). The sensing accuracy
of TDR can only be ensured when the rise time of the drive signal is shorter than 200 ps.
Such technical restriction in rise time makes it difficult to apply the TDR method to soil profile
moisture monitoring [27–29]. Complex secondary calibration of FDR is required before use since it
is subject to the effects of various factors, such as soil type, conductivity, and volumetric weight,
when measuring frequency is low [17,30]. With simpler operation and higher efficiency than FDR,
which requires frequency domain reflection, HFC is considered the best substitute for TDR and is able
to satisfy the demands for real-time online measurement since it is not affected by soil texture in the
HFC field [31,32].
Previous studies on soil profile moisture sensors have revealed various problems, such as high
cost, high data redundancy, and complex operation [7,33–39]. In this study, a soil profile moisture
sensor was designed following the principle of HFC, based on the research achievements on an efficient
soil profile moisture sensing method developed by the National Engineering and Technology Center for
Information Agriculture, Nanjing Agricultural University. With this device, not only could volumetric
moisture content of the 20, 30, and 50 cm soil layers be detected, but online inversion of moisture for
0–100 cm soil depth could also be conducted.

2. Design of the Soil Profile Moisture Sensor

2.1. The Sensing Principle

In the HFC method, soil moisture is obtained by measuring the equivalent capacitance of the
soil mix (water, air, and soil particles) that serves as a medium between metal electrodes at a high
frequency. Soil can be classified into sandy, clay, and loam soils by particle size. Dielectric constant of
soil can vary significantly with different particle sizes and air spaces. However, HFC is immune to soil
 Sensors 2018, 18, 1648 3 of 23
Sensors 2018, 18, x 3 of 24
Sensors 2018, 18, x 3 of 24
soil texture [31]. According to Hilhorst [40], dielectric constants of three types of soils were almost
texture [31]. [31].
According to Hilhorst [40], dielectric constants of threeoftypes of types
soils were almost thealmost
same
thesoil texture
same when theAccording
frequency to of Hilhorst
external [40], dielectric
electric field wasconstants
100–500 MHz three and theofinfluence
soils were of soil
when
the samethe frequency
when of external
the frequency electric field was 100–500 MHz and the influence of soil texture on the
texture on the dielectric constantofwas external electricThe
negligible. field was 100–500
capacitance MHz
of the and the formed
capacitor influence byofthesoil
dielectric
texture onconstant was negligible.
the dielectric constant The capacitance ofThe
the capacitance
capacitor formed bycapacitor
the metalformed
electrodes and
metal electrodes and soil is almostwas negligible.
entirely determined by the soilofmoisture
the content sinceby thethe
soil is almost entirely determined by the soil moisture content since the dielectric constants of solid
dielectric constants of solid matters and gases are far smaller than that of water in soil. Therefore, soilthe
metal electrodes and soil is almost entirely determined by the soil moisture content since
matters
dielectric and gases are
constants far smaller
of solid matters than
and that
gasesofare
water in soil. Therefore,
than that ofsoil
far capacitance.
smaller watermoisture
in soil.content
Therefore,cansoil
be
moisture content can be estimated from the measured
estimated
moisture from the
content measured
can be capacitance.
In order to prevent theestimated
sensor from from the measured
being affected bycapacitance.
the soil corrosion, the annular electrode of
In
In order
order to
to prevent
prevent the
the sensor
sensor from
from being
being affected
affected by
by the
the soil
soil corrosion,
corrosion, the
the annular electrode
annularrings
electrode of
the sensor was constructed as shown in Figure 1a. Two identical metal cylindrical wereof
the
the sensor
sensor waswas constructed
constructed as shown
as and
shownin Figure 1a. Two identical metal cylindrical rings were sheathed
sheathed on the insulating sleeve, the in Figure
outer layer1a.of Two identical
the sensor was metal
covered cylindrical rings
with an insulating were
on the
sheathed insulating sleeve, and the outer layer of the sensor was covered with an insulating tube layer to
tube layer toon the insulating
obstruct the soil sleeve,
corrosion. and the outer layer of the sensor was covered with an insulating
obstruct the soil corrosion.
tube layer to obstruct the soil corrosion.

(a) (b)
(a) (b)
Figure
Figure 1. 1.
Structural diagram
Structural of of
diagram thethe
annular electrode:
annular (a)(a)
electrode: The annular
The electrode;
annular (b)(b)
electrode; The radiation
The radiation
Figure
electric 1. Structural diagram of the annular electrode: (a) The annular electrode; (b) The radiation
field
electric field
electric field
As shown in Figure 1b, when the two metal annular electrodes were connected to the circuit
As shown
shown in Figure
Figure1b,1b,when the two metal annular electrodes were connected to the circuit and
and radiating theinelectric field when
around the two
the metal annular
periphery, electrodes
the measured were
soil connected
and the two to the circuit
electrodes
radiating the
and radiating electric field
the electric around the
field field
around periphery, the measured
the periphery, soil and
the measured the two electrodes
soiltogether.
and the The covered
two physical by
electrodes
covered by the radiation electric constituted an equivalent capacitor
the radiation
covered electric field constituted anconstituted
equivalent capacitor together. The physical structure of the
structure ofbythethe radiation
annular electric
electrode wasfield
shown in Figurean
2. equivalent capacitor together. The physical
annular electrode was shown in Figure 2.
structure of the annular electrode was shown in Figure 2.

Figure 2. The physical structure of the annular electrode.

Figure 2.
Figure 2. The
The physical
physical structure of the
structure of the annular
annular electrode.
electrode.
In Figure 2, D was the axial distance between two electrodes. Because the probe is based on the
In Figure
fringe field 2,toDsense
effect2, was the axial
the axial distance
capacitive between
resistance two
oftwo electrodes. Because
the water-containing the probeisisvery
soil, based on the
In Figure D was the distance between electrodes. Because thewhich
probe is basedsimilar
on the
fringe
tofringe field
the working effect to
principlesense the
of the capacitive
the capacitive resistance
planar capacitance of the water-containing soil, which is very similar
field effect to sense resistancesensor. Therefore, the central
of the water-containing angle isθ very
soil, which was similar
1° in
to paper,
this the working
and the principle
annularofelectrode
the planarwascapacitance
divided intosensor. Therefore,
360 parts, and θthe
forcentral angleand
each upper θ was 1° in
lower
this paper, and the annular electrode was divided into 360 parts, and θ for each upper and lower
Sensors 2018, 18, 1648 4 of 23

to theSensors
working principle
2018,18,
18,xx of the planar capacitance sensor. Therefore, the central angle θ was 1◦24in
4 of
Sensors 2018, 4 of 24
this paper, and the annular electrode was divided into 360 parts, and θ for each upper and lower
annular electrode ◦1°.
isisis11°. Thecorresponding
correspondinglongitudinal
longitudinal
annular
annular electrode
electrode . The
The corresponding longitudinalsection
sectionconstituted
section constitutedaaamicro-quantized
constituted micro-quantized
micro-quantizedplanar
planar
planar
capacitance,
capacitance, as as shownininFigure
shown Figure3.3.
capacitance, as shown in Figure 3.

Figure 3. Diagram of parallel plate capacitance expanding to fringing field capacitance.

Figure 3. Diagram of parallel plate capacitance expanding to fringing field capacitance.
Figure 3. Diagram of parallel plate capacitance expanding to fringing field capacitance.
As shown in Figure 3, the equivalent capacitance can be considered as parallel plate capacitors
As
Asshown
formed shown ininFigure
in parallel Figure
by an3,3,infinite
the
theequivalent
number capacitance
equivalent capacitance
of can
canbe
micro-dimension beconsidered
considered asasparallel
fringing field parallelplate capacitors
plateAfter
capacitances. capacitors
the
formed
formed in
in parallel
parallel by
by an
aninfinite
infinite number
number of
ofmicro-dimension
micro-dimension fringing
fringingfield
field capacitances.
capacitances.
annular probe is micro-divided, the entire probe sensing the capacitive resistance represented by the After
After the
the
annular
annular probe
probe
soil based is micro-divided,
onisthemicro-divided,
high-frequency the entire
the entireprobe
probe
fringing sensing
sensing
electric the capacitive
fieldtheeffect
capacitiveresistance
can beresistance represented
equivalentrepresented by the
by the
to 360 micro
soil
soilbased
basedonin
capacitors the
on high-frequency
the as shown fringing
high-frequency
parallel, electric
fringing
in Figure field effect
4. electric field can be equivalent
effect to 360 micro
can be equivalent to capacitors
360 micro
incapacitors
parallel, as
in shown
parallel,inas
Figure
shown 4. in Figure 4.

Figure 4. The annular probes sensing the equivalent capacitive reactance represented by soil.

Figure 4. The annular probes sensing the equivalent capacitive reactance represented by soil.
4. The annular probes sensing the equivalent planar
capacitive reactance represented
Figure
The capacitance of the probe’s micro-quantized capacitance shown in by soil. 4 was
Figure
calculated using an approximate analytical algorithm, based on the relevant research of the fringing
The capacitance of the probe’s micro-quantized planar capacitance shown in Figure 4 was
field
Thecapacitor [41]. of
capacitance In the
the probe’s
solution-like algorithm, theplanar
micro-quantized unit integration
capacitancemethod
shownwas taken as
in Figure an
4 was
calculated using an approximate analytical algorithm, based on the relevant research of the fringing
example. Generally, the power line between two electrodes was considered as
calculated using an approximate analytical algorithm, based on the relevant research of the fringingan arc curve or an
field capacitor [41]. In the solution-like algorithm, the unit integration method was taken as an
elliptic
field curve[41].
capacitor [42].InThe
theapproximated solution ofthe
solution-like algorithm, a single micro-quantized
unit integration methodcapacitance
was takenofasan
anannular
example.
example.
electrode Generally,
was shown the
in power line
equation (1): between two electrodes was considered as an arc curve or an
Generally, the power line between two electrodes was considered as an arc curve or an elliptic curve [42].
elliptic curve [42]. The approximated solution of a single micro-quantized capacitance of an annular
electrode was shown in equation (1): k0 Rθi (ε r + ε 0 ) 2W of an annular electrode was shown
The approximated solution of a single micro-quantized capacitance
in equation (1): Ci = ln(1 + ) (1)
( επ
Rθi i (rε r +0ε 0ln
kk00Rθ + ε ) D
) (1 + 2W)
2W
CCi == ln(1 +D ) (1)
(1)
i
Among them, π
π D
Among
C i —— them,
Single micro-quantized capacitance
Among them,
k —— Correction
Ci —— Single micro-quantized
0 constant capacitance
CRi —— —— Radius
Single micro-quantized
of the annular capacitance
electrode
k0 —— Correction constant
k0ε—— r ——
Correction
Dielectricconstant of water-containing soil
R —— Radius of theconstant
annular electrode
Rε—— Radius of the annular electrode
ε r —— 0 ——Dielectric constant
Composite of water-containing
dielectric constant of othersoil radiation-field media
ε r —— Dielectric constant of water-containing soil
ε 0 θ—— Composite dielectric constant of other radiation-field media
i —— The central angle of the annular electrode corresponding to the micro-quantized
ε ——0 —— The centralisangle of the constant
annular
θ
fringing
i field,Composite
which dielectric
made constant by 1° electrode
of=other
(i corresponding
1, 2, 3radiation-field
... 360) to the micro-quantized fringing
media
◦
θW
field, which
i ———— is Width
made constant
The central by 1 of
angle
of the annular (i =the
1, 2,annular
electrode 3 ... 360)electrode corresponding to the micro-quantized
W D —— —— Width
Spacingof the
of annular
annular electrode
electrodes
fringing field, which is made constant by 1° (i = 1, 2, 3 ... 360)
DW—— ——Spacing
Width ofof the
annular electrodes
annular electrode
D —— Spacing of annular electrodes
Sensors Sensors
2018, 18,2018,
164818, x 5 of
5 of 23 24

In the case of non-changed structural parameters (the annular electrode width W, spacing D
and Inradius
the case R),ofthe single micro-quantized
non-changed capacitance
structural parameters (thevalue C i electrode
annular to theW,dielectric
is relatedwidth D and
spacing constant
( ε r + ε 0 ), so the dielectric constant of the measured soil can be converted to measure the size of the
radius R), the single micro-quantized capacitance value Ci is related to the dielectric constant (ε r + ε 0 ),
so the dielectric constant of the measured soil can be converted to measure the size of the capacitance
capacitance represented by the electric field.
represented by the electric field.
An easy way to measure capacitance is to place the equivalent capacitance characterized by the
An easy way to measure capacitance is to place the equivalent capacitance characterized
electrodes and soil into an LC oscillating circuit and measure the final circuit frequency. The
by the electrodes and soil into an LC oscillating circuit and measure the final circuit frequency.
equivalent circuit is shown in Figure 5. The voltage-controlled oscillator generates a high-frequency
The equivalent circuit is shown in Figure 5. The voltage-controlled oscillator generates a high-frequency
electromagnetic fringe field effect by means of fixed resonant capacitance and an inductor. The
electromagnetic fringe field effect by means of fixed resonant capacitance and an inductor. The capacity
capacity of soil’s equivalent capacitance that is detected by the sensor’s probe is associated with the
of soil’s equivalent capacitance that is detected by the sensor’s probe is associated with the stray
stray capacitance of surrounding soil and the probe itself [41]. Formed by this composite structure,
capacitance of surrounding soil and the probe itself [41]. Formed by this composite structure, this probe
this probe is equivalent to an equivalent capacitance with a stray capacitance. In order to allow the
is equivalent to an equivalent capacitance with a stray capacitance. In order to allow the microprocessor
microprocessor to directly measure the output frequency, it is necessary to input the high-frequency
to directly measure the output frequency, it is necessary to input the high-frequency signal into the
signal into the signal processing frequency divider circuit, so that the signal is converted into a
signal processing frequency divider circuit, so that the signal is converted into a low-frequency
low-frequency signal output.
signal output.

Figure 5. Equivalent
Figure circuit
5. Equivalent of of
circuit thethe
high-frequency capacitance
high-frequency method.
capacitance method.

Equation
Equation (3)(3) indicates
indicates thatthat
thethe frequency
frequency generated
generated by aby a voltage-controlled
voltage-controlled oscillator
oscillator is subject
is subject to
thetonominal
the nominal
values values of capacitance
of capacitance and inductors
and inductors in an LCcircuit.
in an LC resonance resonance circuit. The
The dielectric dielectric
constant of
constant
soil of soil
under test under
varies testsoil
as the varies as the content
moisture soil moisture content
changes, changes,
which results which results
in changes of in
itschanges of its
equivalent
equivalentand
capacitance capacitance and circuit’s
circuit’s output output frequency.
frequency.

f = p 1 1
f = (2)
(2)
2π2π L(L
Co(Co + Cp
+ Cp + )Cs )
+ Cs

where
wheref isf is
thethe
frequency;
frequency; Co,Co,
Cs,Cs,
and CpCp
and areare
thethe
fixed resonant
fixed resonant capacitance,
capacitance,equivalent
equivalent capacitance,
capacitance,
and stray
and stray capacitance
capacitance in in
thethe
circuit, respectively;
circuit, respectively; and L is
and L the inductor.
is the inductor.The
Thedielectric constant
dielectric soil
constant soil
characterized
characterized varies
varieswhen
when soil moisture
soil moisturecontent
contentchanges,
changes, followed
followed byby
thethe
equivalent
equivalentcapacitance
capacitancethethe
probe
probe detected,
detected, output
output frequency
frequency of high-frequency
of high-frequency oscillator,
oscillator, andDC
and the thevoltage
DC voltage signals
signals sent
sent by theby
the sensor
sensor after conditioning.
after signal signal conditioning.
Thus, theThus, the soil moisture
soil moisture content can content can bebased
be inverted inverted based
on the on the
changing
DCchanging DC voltage signals.
voltage signals.

2.2.2.2.
Structural Design
Structural of the
Design Sensor
of the Sensor
AsAsshown
shownin Figure 6a, the
in Figure 6a,sensor consisted
the sensor of five sensing
consisted of five probes,
sensinga probes,
data processor,
a data and accessory
processor, and
components. The five groups
accessory components. The of sensing
five groups probes were capacitors
of sensing probes were comprising
capacitorstwo annular electrodes.
comprising two annular
These five probes
electrodes. Thesewere
five designed
probes were to measure
designedthe water layer,
to measure air layer,
the water andairatlayer,
layer, soil depths
and at of
soil20, 30,
depths
and 50 cm. 20, 30 and 50 cm represents the depth position of the sensor-sensitive probe
of 20, 30, and 50 cm. 20, 30 and 50 cm represents the depth position of the sensor-sensitive probe in in the soil,
respectively. And the length
the soil, respectively. Andof each
the layer
length ofiseach
100 layer
mm. Capacitors
is 100 mm.for the air and
Capacitors forwater
the airlayers were able
and water layers
to were
dynamically calibrate each sensor.
able to dynamically calibrate each sensor.
Sensors 2018, 18, 1648 6 of 23
Sensors 2018, 18, x 6 of 24

(a) (b)
Figure
Figure 6. (a)6.Composition
(a) Composition diagram
diagram and (b)and (b) physical
physical productproduct
of the soilofprofile
the soil profilesensor;
moisture moisture
sensor;
1 Data
① Data
processor; processor; ② Water-layer
2 Water-layer capacitor; capacitor; ③ Outer wall
3 Outer wall of the sensor; of the sensor; ④ Annular
4 Annular electrode; electrode;
5 Sensing
⑤ Sensing
probe;
6 Insulation ⑥ Insulation sleeve.
probe;sleeve.

The sensor was sealed using acrylic organic glasses. The five sensing probes were fixed to the
The sensor was sealed using acrylic organic glasses. The five sensing probes were fixed to the
insulation sleeve. The sensing probes and the data processor were connected to the power supply by
insulation sleeve. The sensing probes and the data processor were connected to the power supply by
shielded wires. At the bottom of the data processor, there were multiple sets of signal terminals used
shielded wires. At the bottom of the data processor, there were multiple sets of signal terminals used
for signal frequencies outputted by various sensing probes. Signal interference and overpower
for signal frequencies outputted by various sensing probes. Signal interference and overpower during
during simultaneous function of multiple groups of sensing probes were avoided by time-sharing of
simultaneous function of multiple groups of sensing probes were avoided by time-sharing of the
the power supply. In order to avoid interference among electric fields and power overloading
power supply. In order to avoid interference among electric fields and power overloading problems
problems between adjacent sensor probes when the soil moisture content information is measured at
between adjacent sensor probes when the soil moisture content information is measured at all levels at
all levels at the same time, five groups of sensor sensitive probes are operated by time-sharing
the same time, five groups of sensor sensitive probes are operated by time-sharing control when the
control when the sensor is working. The solid-state relays (optical coupling isolation relay) of type
sensor is working. The solid-state relays (optical coupling isolation relay) of type AQY212S are used.
AQY212S are used. The control port of the microprocessor emits high and low levels to drive the size
The control port of the microprocessor emits high and low levels to drive the size of the output current,
of the output current, and large or small currents are fed into the front end of the solid state relay to
and large or small currents are fed into the front end of the solid state relay to control the connection
control the connection and disconnection between two SWITCH ports at the rear end, respectively.
and disconnection between two SWITCH ports at the rear end, respectively. The typical value of the
The typical value of the relay turn-on time is 0.65 ms, and the closing time is only 0.08 ms.
relay turn-on time is 0.65 ms, and the closing time is only 0.08 ms.
2.3. Sensing Probe Design
2.3. Sensing Probe Design

2.3.1.
2.3.1. Simulation
Simulation Model
Model of the
of the Sensing
Sensing Probe’s
Probe’s Annular
Annular Electrode
Electrode
TheThe electric
electric field
field distribution
distribution around
around thethe annular
annular electrode
electrode waswas determined
determined by the
by the electrode’s
electrode’s
structural characteristics, which further affected the sensitivity of capacitance detection
structural characteristics, which further affected the sensitivity of capacitance detection and range and range of
detection. According
of detection. According to toEquation
Equation(1),
(1),the
thesize
sizeofof
thethe equivalent
equivalent capacitance
capacitance formed
formed by by
thethe sensor
sensor
probe adopting the annular electrode structure and the soil is affected
probe adopting the annular electrode structure and the soil is affected not only by the vacuumnot only by the vacuum
dielectric
dielectric constant
constant and
and the the relative
relative dielectricdielectric
constant, constant,
but also bybutthe also by thestructural
capacitance capacitance structural
parameters
(the annular electrode width W, the spacing D and radius R). Therefore, simulation of the annular of
parameters (the annular electrode width W, the spacing D and radius R). Therefore, simulation
the annular
electrode’s electrode’s
different geometricdifferent geometric
sizes was required.sizes
Thewas required.
geometric The
sizes geometric
of the annularsizes of theofannular
electrodes the
electrodes
sensing probesofwere
the identified
sensing probes were
based on theidentified
distributionbased on thefield
of electric distribution of electric
around them. fieldwidth
Electrode around
andthem. Electrode
spacing widthasand
are denoted W spacing
and D. are denoted as W and D.
A high-frequency structure simulator (HFSS) from ANSYS Inc was used to simulate the
radiation intensity of electric field generated by the annular electrodes and thus determine the
 Sensors 2018, 18, 1648 7 of 23

A high-frequency structure simulator (HFSS) from ANSYS Inc was used to simulate the radiation
Sensors 2018, 18, x 7 of 24
intensity of electric field generated by the annular electrodes and thus determine the optimal structure
of the structure
optimal annular electrodes. A simulation
of the annular model
electrodes. of the annular
A simulation model electrode structureelectrode
of the annular was built.structure
As shown
was built. As shown in Figure 7, the external cylinder was soil space and the internal yellow ring whose
in Figure 7, the external cylinder was soil space and the internal yellow ring was a copper ring was
a inner
copper andring
outer diameters
whose innerwere
anduniformly set to 48 mm
outer diameters wereanduniformly
51 mm, respectively.
set to 48 Suppose
mm and the51
boundary
mm,
of electric field
respectively. Supposegenerated by the annular
the boundary electrodes
of electric was ideal; by
field generated thethe
columnar
annularsoil space that
electrodes was
was 40 cm
ideal;
the columnar soil space that was 40 cm in diameter and 30 cm in height was the boundary condition of
in diameter and 30 cm in height was the boundary condition of radiation; the dielectric constants
of the packingthe
radiation; medium (soilconstants
dielectric volumetric of moisture
the packing content was 34%)
medium (soil [41], the acrylic
volumetric installation
moisture contentpipe and
was
the shaft in the pipe, and air in the pipe were 21, 3.5, and 1, respectively; the solving
34%) [41], the acrylic installation pipe and the shaft in the pipe, and air in the pipe were 21, 3.5, and 1, frequency was
150 MHz, lumped
respectively; port frequency
the solving stimulation;wasW 150
wasMHz,
set to lumped
10, 20, and
port30stimulation;
mm; copperW ring
was spacing was20,
set to 10, setand
to 10,
30 20,
mm;30,copper
40, 50, ring
and spacing
60 mm. was set to 10, 20, 30, 40, 50, and 60 mm.

Figure 7. Simulation model of the annular electrode structure.

Figure 7. Simulation model of the annular electrode structure.

2.3.2. Simulation of the Sensing Probe’s Annular Electrode

2.3.2. Simulation of the Sensing Probe’s Annular Electrode
The electric field distribution and shape obtained by the annular electrode simulation with
The electric field distribution and shape obtained by the annular electrode simulation with
different structure parameters varied. The effective lateral radiation range was determined by the
different structure parameters varied. The effective lateral radiation range was determined by the
diameter of the annular electrode. The effective radial radiation range and contiguity of the electric
diameter of the annular electrode. The effective radial radiation range and contiguity of the electric
field were determined by the spacing between two annular electrodes. The results of the annular
field were determined by the spacing between two annular electrodes. The results of the annular
electrode simulation with different structure parameters are shown in Table 1. According to Table 1,
electrode simulation with different structure parameters are shown in Table 1. According to Table 1,
when copper ring width (W) was 10 mm, the largest radiation range was observed when copper ring
when copper ring width (W) was 10 mm, the largest radiation range was observed when copper ring
spacing (D) was 60 mm and axial and radial radiations were 100.00 mm and 80.00 mm, respectively.
spacing (D) was 60 mm and axial and radial radiations were 100.00 mm and 80.00 mm, respectively.
When W was 20 mm, the largest axial radiation range was observed when D was 40 mm and axial
When W was 20 mm, the largest axial radiation range was observed when D was 40 mm and axial
and radial radiations were 106.00 mm and 80.00 mm, respectively. The largest radial radiation range
and radial radiations were 106.00 mm and 80.00 mm, respectively. The largest radial radiation range
was observed when D was 60 mm and axial and radial radiations were 98.00 mm and 100.00 mm,
was observed when D was 60 mm and axial and radial radiations were 98.00 mm and 100.00 mm,
respectively. When W was 30 mm, the largest axial radiation range was observed when D was 40
respectively. When W was 30 mm, the largest axial radiation range was observed when D was
mm and axial and radial radiations were 94.50 mm and 100.00 mm, respectively; the largest radial
40 mm and axial and radial radiations were 94.50 mm and 100.00 mm, respectively; the largest radial
radiation range was observed when D was 60 mm and axial and radial radiations were 90.00 mm
radiation range was observed when D was 60 mm and axial and radial radiations were 90.00 mm and
and 120.00 mm, respectively.
120.00 mm, respectively.
A comparison
Table 1. Simulationofresult
simulation results
of annular of three
electrodes structure
with differentparameters with the largest
structure parameters (unit of distance
radiationof axial
radiation shown in Figure 8. When W was 10 mm and D was 60 mm, the radial radiation diameter
distance:ismm).
was 100.00 mm with uniform distribution of field intensity, but the range of radial plane radiation
Copper Ring Spacing (D)
was only 80.00Widthmm.
(W) When W was 20 mm and D was 40 mm, axial and radial field intensity was not
Direction
10 mm 20 mm 30 mm 40 mm 50 mm 60 mm
uniformly distributed, the surrounding
Axial 82.00
electric field was
90.00 91.00
not compact,
94.50
axial
100.00
radiation
100.00
diameter was
10 mmmm, and the range of radial radiation was 80.00 mm. When W was 30 mm and D
as high as 106.00 Radial 30.00 40.00 50.00 60.00 70.00 80.00
was 40 mm, the axial
20 mm
and radial
Axial field intensity
84.00 was
95.00 uniformly
95.20 distributed;
106.00 the
98.00 axial radiation diameter
98.00
was 94.50 mm, and the range Radial of radial
50.00radiation
60.00was 100.00
70.00 mm. 80.00 90.00 considered,
All things 100.00 the spacing
Axial 85.00 92.00 94.00 94.50 92.50 90.00
of sensitive soil layer-sensing for the sensor was set to 100 mm, as well as the radiation range and
30 mm
Radial 70.00 80.00 90.00 100.00 110.00 120.00

A comparison of simulation results of three structure parameters with the largest distance of
axial radiation is shown in Figure 8. When W was 10 mm and D was 60 mm, the radial radiation
diameter was 100.00 mm with uniform distribution of field intensity, but the range of radial plane
Sensors 2018, 18, 1648 8 of 23

continuity of the electric field, W = 30 mm and D = 40 mm were taken as the optimal structure of the
sensor probe’s annular electrode.

Table 1. Simulation result of annular electrodes with different structure parameters (unit of radiation
distance: mm).
Sensors 2018, 18, x 8 of 24
Copper Ring Spacing (D)
Width (W) Direction
radiation was only 80.00 mm. When 10 W mm
was 20 mm
20 mmand D30was
mm40 mm,
40 mmaxial and radial 60
50 mm field
mmintensity
was not uniformly distributed, the surrounding electric field was not compact, axial radiation
Axial 82.00 90.00 91.00 94.50 100.00 100.00
10 mm
diameter was as high asRadial
106.00 mm, 30.00
and the range
40.00of radial radiation
50.00 was 80.00
60.00 70.00mm. When
80.00 W was
30 mm and D was 40 mm, the axial and radial field intensity was uniformly distributed; the axial
Axial 84.00 95.00 95.20 106.00 98.00 98.00
20 mm was 94.50 mm, and the range of radial radiation was 100.00 mm. All things
radiation diameter Radial 50.00 60.00 70.00 80.00 90.00 100.00
considered, the spacing of sensitive soil layer-sensing for the sensor was set to 100 mm, as well as the
Axial 85.00 92.00 94.00 94.50 92.50 90.00
radiation range
30 mmand continuity of the electric field, W = 30 mm and D = 40 mm were taken as the
Radial 70.00 80.00 90.00 100.00 110.00 120.00
optimal structure of the sensor probe’s annular electrode.

(a) (b) (c)

Figure 8. Comparison between the three combinations that have the largest axial radiation distance
Figure 8. Comparison between the three combinations that have the largest axial radiation distance
(a) W = 10 mm and D = 60 mm, (b) W = 20 mm and D = 40 mm, and (c) W = 30 mm and D = 40 mm.
(a) W = 10 mm and D = 60 mm, (b) W = 20 mm and D = 40 mm, and (c) W = 30 mm and D = 40 mm.

2.4.
2.4.Circuit
CircuitDesign
Design
The
Thedesign
designprinciples
principles of of the circuit of
the circuit of the
thesoil
soilprofile
profilemoisture
moisture sensor
sensor areare
shown shown in Figure
in Figure 9. 9.
It Itfunctionally consisted of two parts, the sensor acquisition circuit, and
functionally consisted of two parts, the sensor acquisition circuit, and the control circuit. The the control circuit.
The acquisition
acquisition circuit
circuit comprises
comprises annular
annular metal
metal electrodes,
electrodes, voltage-controlled
voltage-controlled oscillating
oscillating circuit,
circuit,
amplitude
amplitudeamplifying
amplifyingcircuit,
circuit, and
and frequency dividing/conditioning
frequency dividing/conditioning circuit,
circuit, and
and it isit used
is used to monitor
to monitor
soil
soilmoisture
moisturecontent.
content. The sensor control circuit
sensor control circuitcomprises
comprisesananATmega328p
ATmega328p 8-bit
8-bit microcontroller
microcontroller
(MCU),
(MCU), thethe power
power circuit,
circuit, the the
485 485
cablecable communication
communication and transmission
and transmission module,
module, NRF 24L01
NRF 24L01 wireless
2.4wireless 2.4 G communication
G communication and transmission
and transmission module,module, and time-sharing
and time-sharing powerpower
supplysupply circuit
circuit and and
power
powerThe
circuit. circuit.
sensorThewassensor was powered
powered by a 7.4 V bylithium
a 7.4 Vbattery
lithiumandbattery and was to
was accessed accessed
the solar to module.
the solar
module.
Among them, the voltage-controlled oscillating circuit is different from the voltage-controlled
Among
oscillation modethem, the voltage-controlled
application circuit commonly oscillating
used in circuit
the VCO is different
chip. An from the voltage-controlled
LC resonant mode application
oscillation mode application circuit commonly used in the VCO chip.
circuit of the voltage-controlled oscillation chip is used to place the equivalent capacitance An LC resonantformed mode by
theapplication circuit of
annular electrode andthesoilvoltage-controlled
in the LC oscillation oscillation
circuit, sochip
as toisgenerate
used toa place
variable thehigh-frequency
equivalent
capacitance formed by the annular electrode and soil in the LC oscillation circuit, so as to generate a
variable high-frequency sinusoidal signal. The voltage-controlled oscillation chip uses the MC1648d
from Freescale Company (currently NXP). The principle of voltage-controlled oscillator and LC
resonance circuit is shown in Figure 10. An external LC and annular electrodes were arranged in
parallel between Pin 1 (Tank) and Pin 8 (Bias) on the chip to make up the resonance tank, using the
high-frequency oscillating circuit that outputs the oscillation frequency with a variable equivalent
 Sensors 2018, 18, 1648 9 of 23

Sensors 2018,
sinusoidal 18, x The voltage-controlled oscillation chip uses the MC1648d from Freescale Company
signal. 9 of 24
(currently NXP). The principle of voltage-controlled oscillator and LC resonance circuit is shown in
capacitance.
Figure 10. An As shown
externalinLC Figure
and 10, the inductor
annular electrodes L1 is encapsulated
were arranged in byparallel
the patch 0805 with
between Pinthe size
1 (Tank)
ofand
0.1 μH order of magnitude. Both C2 and C3 are 100 pF capacitors
Pin 8 (Bias) on the chip to make up the resonance tank, using the high-frequency oscillating encapsulated by the patch 0805,
the connecting
circuit terminal
that outputs theP1 is used tofrequency
oscillation connect two with brass annularequivalent
a variable electrodes,capacitance.
the three of C2, C3, andin
As shown
P1 constitute the
Figure 10, theSensors capacitance
inductor compensation circuit of the soil equivalent
L1 is encapsulated by the patch 0805 with the size of 0.1 µH order9 of capacitance to avoid the
of24magnitude.
2018, 18, x
input multivalued problem. Capacitor C4 and C15 use variable capacitors
Both C2 and C3 are 100 pF capacitors encapsulated by the patch 0805, the connecting terminal P1 is in the order of several tens
ofused
pF asto resonant
capacitance.
connect capacitors
two As shown
brass by adjusting
in Figure
annular 10, thethe size
inductor
electrodes, theL1 ofisthe
three ofvariable
C2, C3,bycapacitor.
encapsulated the patch
and Onwith
0805
P1 constitute the the
one
the hand, the
size
capacitance
sensitivity of 0.1the
and μH range
order ofof magnitude.
the soil Both C2 and C3
equivalent are 100 pF capacitors
capacitance to the encapsulated
output by the patch
frequency are 0805,
adjusted. On
compensation circuit of the soil equivalent capacitance to avoid the input multivalued problem.
the connecting terminal P1 is used to connect two brass annular electrodes, the three of C2, C3, and
the other
Capacitor P1hand, the
C4constitute
and C15theinherent
use error caused
variable compensation by
capacitors in the structure
the order and
of several component
tens accuracy
of pF astoresonant of multiple
capacitance circuit of the soil equivalent capacitance avoid thecapacitors
sensor circuits
by adjusting input are
the also adjusted,
size of the
multivalued so
variable
problem. as to ensure
capacitor.
Capacitor C4 and On the consistency
thevariable
C15 use one hand, of the output
the sensitivity
capacitors of each
in the order ofand sensor.
thetens
several C1,ofC5,
range the
C6, and C7 are
of pF all
as patch
resonant capacitors
capacitors encapsulated
by adjusting the by
size 0805,
of the where
variable C6 is different
capacitor.
soil equivalent capacitance to the output frequency are adjusted. On the other hand, the inherent error On the from
one C1,
hand, C5,
the and C7
incaused
order by ofsensitivity
magnitude,
the
and theC5
structure and
rangeandof the
C6soilare
component
equivalent
used capacitance
accuracy
to the output
asofdecoupling
multiple sensor
frequency
capacitors arefilter
circuitstoare
adjusted.
alsoout
On high and
adjusted, so as
the other hand, the inherent error caused by the structure and component accuracy of multiple
low-frequency
to ensure sensor noise
the consistencyinterference from
of adjusted,
the output the power supply and the ground plane, respectively. C1 and
circuits are also so as of each sensor.
to ensure C1, C5,
the consistency C6,
of the andofC7
output are
each all patch
sensor. C1, C5,capacitors
C7 are usedC6,
encapsulated asand
filter
by 0805,capacitors
C7 are where
all patchC6 toisfilter out
different
capacitors the
from
encapsulatedinterference
C1, C5, and
by 0805, ofC7
where parasitic
C6inis order impedance
differentoffrom
magnitude, to and
C1, C5, the
C5C7oscillating
and C6 are
loop in high-frequency
in order of circuits.
magnitude, C5 and C6 are used as decoupling capacitors
used as decoupling capacitors to filter out high and low-frequency noise interference from the power to filter out high and
low-frequency noise interference from the power supply and the ground plane, respectively. C1 and
supply and the ground plane, respectively. C1 and C7 are used as filter capacitors to filter out the
C7 are used as filter capacitors to filter out the interference of parasitic impedance to the oscillating
interference ofinparasitic
loop high-frequencyimpedance
circuits. to the oscillating loop in high-frequency circuits.

Figure 9. Design principles of the circuit of the soil profile moisture sensor.
Design
Figure9.9.Design
Figure principles
principles ofof thecircuit
the circuitofofthe
thesoil
soilprofile
profilemoisture
moisturesensor.
sensor.

Figure 10. Schematic diagram of the LC resonance circuit.

Signal processing primarily refers to the frequency division, conditioning, and shaping of
frequency signals, i.e., conducting frequency division of original frequency signals to allow the MCU
to measure them directly. The change of soil moisture content leads to the variability of the

Figure10.
Figure Schematicdiagram
10.Schematic diagramofofthe
theLC
LCresonance
resonancecircuit.
circuit.

Signal processing primarily refers to the frequency division, conditioning, and shaping of
frequency signals, i.e., conducting frequency division of original frequency signals to allow the MCU
to measure them directly. The change of soil moisture content leads to the variability of the
Sensors 2018, 18, 1648 10 of 23

Signal processing primarily refers to the frequency division, conditioning, and shaping of
frequency signals, i.e., conducting frequency division of original frequency signals to allow the
MCU to measure them directly. The change of soil moisture content leads to the variability of the
equivalent capacitance within a certain range. The output high frequency actually measured by the
voltage-controlled oscillating circuit is between 100 MHz and 150 MHz. To enable the measured
outputSensors
frequency
2018, 18, xcan be read by the microprocessor, it is necessary to input high-frequency 10 of 24 signals
into the signal processing frequency divider circuit and also input transfer high-frequency signals
equivalent capacitance within a certain range. The output high frequency actually measured by the
into low-frequency signals. In this study, the original frequency signals were divided by 64 using
voltage-controlled oscillating circuit is between 100 MHz and 150 MHz. To enable the measured
the output
dual-mode pre-scaler
frequency andby
can be read then divided by 256it using
the microprocessor, a SN74HC393D
is necessary digital chip,
to input high-frequency and the result
signals
wasinto
thethe
measurement frequency. To enable the back-end processor to measure this
signal processing frequency divider circuit and also input transfer high-frequency signals frequency signal
into low-frequency signals. In this study, the original frequency signals were divided
directly, MB504 and SN74HC393D typical frequency dividing circuits were used in this study [13]. by 64 using the
dual-mode pre-scaler
SN74HC393D’s digitaland then
chip isdivided
a dual by
four256bit
using a SN74HC393D
binary digital chip,by
counter produced and theThe
TI. result was
maximum count
the measurement frequency. To enable the back-end processor to measure this frequency signal
range is 256. When the 256 signal pulses are input, the output will be flipped and the 256 frequency of
directly, MB504 and SN74HC393D typical frequency dividing circuits were used in this study [13].
the SN74HC393D’s
input frequency will be realized.
digital chip is a dual four bit binary counter produced by TI. The maximum count
range is 256. When the 256 signal pulses are input, the output will be flipped and the 256 frequency
2.5. of
Software
the inputDesign
frequency will be realized.
According to system requirements, the software realized the acquisition of soil profile moisture
2.5. Software Design
data and data processing and transmission. The major function of data acquisition was to acquire the
According to system
output frequencies requirements,
of sensing probes theon software
five groups realized the acquisition
of sensors, of soil profile
i.e., calling moisture
the time-sharing power
data and data processing and transmission. The major function of data acquisition was to acquire the
supply program to switch on the power of the acquisition circuit corresponding to the current loop
output frequencies of sensing probes on five groups of sensors, i.e., calling the time-sharing power
variable
supplyand selecttoto
program openonthe
switch thechannel
power of for
the the input of
acquisition frequency
circuit signals
corresponding to of
thethis acquisition
current loop circuit
to complete data acquisition. Data processing was designed to call the digital
variable and select to open the channel for the input of frequency signals of this acquisition circuit to filtering module to
remove invalid
complete datameasured
acquisition.data
Dataamong multiple
processing frequency
was designed signals,
to call obtain
the digital the final
filtering frequency
module to value,
andremove
convert invalid measured data
the measured among value
frequency multiple frequency
into signals, obtain
the volumetric the final
moisture frequency
content value,
of soil sample under
testand
usingconvert the measured frequency value into the volumetric moisture content of soil sample
the soil moisture content conversion subprogram. The detailed procedure is presented as
under test using the soil moisture content conversion subprogram. The detailed procedure is
followed. Firstly, to make a macro definition on the sample number with a constant N fixed at 101.
presented as followed. Firstly, to make a macro definition on the sample number with a constant N
Secondly, cycle sampling
fixed at 101. Secondly, N times
cycle samplingwith calling
N times thecalling
with measurement function
the measurement everyevery
function 1 ms,1 and
ms, assign the
output
and value
assign theto an array
output space
value to anwith a space
array sizeaof
space with N. Thirdly,
space size of N.all the values
Thirdly, all the in the in
values space
the are sorted
space are sorted
by subscripts frombylargest
subscripts from largest
to smallest. to smallest.
Finally, Finally, the
the program programand
is ejected is ejected
median andvalues
medianof the array
values
space are of the arrayas
returned space
the are returned
filter output as result.
the filterData
output result. Data transmission
transmission was intended was intended
to “pack” to and “split”
“pack” and “split” soil profile moisture after it was acquired and processed, split variables based on
soil profile moisture after it was acquired and processed, split variables based on data size to ensure
data size to ensure smooth data transmission, and guarantee that the data at the receiving end were
smooth
mergeddata andtransmission, and guarantee
restored to obtain the original that the values.
variable data atAthe receiving
software end block
function were diagram
merged isand restored
to obtain
shown the original
in Figure 11. variable values. A software function block diagram is shown in Figure 11.

Figure
Figure11.11.
Block diagram
Block of theof
diagram software functions.
the software functions.

3. Performance Test Design, Data Analysis Methods, and Results

3.1. Performance Tests

 Sensors 2018, 18, 1648 11 of 23

3. Performance Test Design, Data Analysis Methods, and Results

3.1. Performance
Sensors 2018, 18, x Tests 11 of 24

3.1.1.
3.1.1. Laboratory
Laboratory Test
Test
A concentric
A concentric holehole
thethe
sizesize of the
of the sensor’s
sensor’s outer-wall
outer-wall diameter
diameter waswas excavated
excavated on bottom
on the the bottomof a of
a plastic
plastic bucket
bucket withwith an opening
an opening diameter,
diameter, bottom
bottom diameter,
diameter, height,
height, andand volume
volume of 310
of 310 mm, mm,210210
mm,mm,
280 mm, and 12 L, respectively. Next, a 350 mm-long insulation tubular column was inserted into thethe
280 mm, and 12 L, respectively. Next, a 350 mm-long insulation tubular column was inserted into
hole
hole andand fixed
fixed and
and sealed
sealed with
with hot-melt
hot-melt adhesive.
adhesive. The
The sensor
sensor waswas
thenthen inserted
inserted into
into thethe soil
soil samples
samples
where the insulated outer-wall was fixed. The shaft was manually moved up
where the insulated outer-wall was fixed. The shaft was manually moved up and down such that and down such that
sensor probes at different levels could be immersed in the soil. A testing image is shown in Figure 12.12.
sensor probes at different levels could be immersed in the soil. A testing image is shown in Figure

(a) (b)
Figure 12.12.
Figure Photographs of the
Photographs experimental
of the setset
experimental up.up.
(a)Test equipment,
(a)Test (b)(b)
equipment, Operation process.
Operation process.

(1) Calibration of Soil Volumetric Moisture Content

(1) Calibration of Soil Volumetric Moisture Content
Loam soil samples were collected in Rugao, Jiangsu Province, China, with a volumetric weight
Loam soil samples were collected in Rugao, Jiangsu Province, China, with a volumetric weight
of 1.12 g/cm3. In a laboratory with a constant temperature in the range 15–25 °C, the samples were
of 1.12 g/cm3 . In a laboratory with a constant temperature in the range 15–25 ◦ C, the samples were
divided into nine density gradients with the volumetric moisture content from low to high until
divided into nine density gradients with the volumetric moisture content from low to high until nearly
nearly saturated (i.e., 0.00, 0.07, 0.13, 0.20, 0.26, 0.31, 0.37, 0.47, and 0.59; unit: 3m3/m3 3) and then stored
saturated (i.e., 0.00, 0.07, 0.13, 0.20, 0.26, 0.31, 0.37, 0.47, and 0.59; unit: m /m ) and then stored in
in a sealed barrel for 48 h. The soil sample of every gradient was measured 10 times with one of the
a sealed barrel for 48 h. The soil sample of every gradient was measured 10 times with one of the soil
soil profile moisture sensor’s probes (30 cm). The soil prepared to be tested is weighed via the
profile moisture sensor’s probes (30 cm). The soil prepared to be tested is weighed via the electronic
electronic balance scale with the accuracy of 0.01 g. Then, the soil after being weighed will be dried
balance scale with the accuracy of 0.01 g. Then, the soil after being weighed will be dried out via
out via the electrothermal blowing dry box before the dry soil being weighted again. With the
the electrothermal blowing dry box before the dry soil being weighted again. With the oven-drying
oven-drying method, the actual water content of the volume of the tested soil is calculated. The
method, the actual water content of the volume of the tested soil is calculated. The functional
functional relationship between output frequency of acquisition circuit at the sensor’s measuring
relationship between output frequency of acquisition circuit at the sensor’s measuring end and actual
end and actual soil volumetric moisture content was determined. In the same environment, soil
soil
samples volumetric moisture
of five other content
gradients was0.25,
(0.11, determined.
0.32, 0.46,Inand
the0.51;
sameunit:
environment,
m3/m3) were soilmeasured
samples ofwith
five the
other
gradients (0.11, 0.25, 0.32, 0.46, and 0.51; unit: m 3 /m3 ) were measured with the same method. The data
same method. The data were substituted into the functional relationship to test the accuracy of
were substituted
calibration into
of the soil the functional
volumetric relationship
moisture content.to test the accuracy of calibration of the soil volumetric
moisture content.
(2)(2)
Consistency Test
Consistency Test
Test materials
Test materialsand
andenvironment
environment that were
that wereidentical to to
identical those
thoseof of
thethe
soil volumetric
soil volumetric moisture
moisture
monitoring model were used to eliminate individual differences. The five probes were
monitoring model were used to eliminate individual differences. The five probes were inserted inserted into
into
thethe
barrel to to
barrel measure
measure each kind
each kindof of
moisture treatment.
moisture Actual
treatment. Actualsoil volumetric
soil volumetric moisture
moisturecontent was
content was
obtained with
obtained thethe
with oven-drying
oven-drying method.
method.

(3) Verification Test with Different Soil Types

To verify whether the proposed sensor was widely adaptable to different soil types, the soils
used for test were collected from China’s four major agricultural provinces, namely yellow brown
Sensors 2018, 18, 1648 12 of 23

(3) Verification Test with Different Soil Types

To verify whether the proposed sensor was widely adaptable to different soil types, the soils used
for test were
Sensors collected
2018, 18, x from China’s four major agricultural provinces, namely yellow brown soil12from of 24
Jiangsu Province, sandy loam from Jiangxi Province, clay loam from Heilongjiang Province, and sandy
Province,
loam from and
Henan sandy loam from
Province. Henandistribution
Concrete Province. Concrete
of soils distribution
is shown in of soils 13.
Figure is shown in Figure 13.
Soils underwent
Soils underwent
pretreatment, pretreatment,
including including
weed removal, weed removal,
air drying, grinding,air
anddrying, grinding,
sieving, andwere
before they sieving, before
placed in
they were placed in the sealed
the sealed barrel for measurement. barrel for measurement.

Figure13.
Figure 13.Sampling
Samplingregions
regionsand
andtypes
typesofofsoil
soilsamples.
samples.

Thesoil
The soilsamples
samplescollected
collectedfrom
fromJiangxi
JiangxiProvince,
Province,Henan HenanProvince,
Province,and andHeilongjiang
HeilongjiangProvinceProvince
were prepared to different soil volumetric moisture contents using
were prepared to different soil volumetric moisture contents using the same method as above. Density the same method as above.
Density gradients
gradients of soil samplesof soilwere
samples wereamong
variable variable amongsince
samples samples since volumetric
volumetric weight andweight and soil
soil moisture
moisture content of soil samples collected from one region was different
content of soil samples collected from one region was different to those of soil samples collected to those of soil samples
from
collected from another region. The prepared moisture contents of soil
another region. The prepared moisture contents of soil samples from Jiangxi Province were 0.00, 0.10, samples from Jiangxi Province
were
0.14, 0.00,
0.21, 0.10,
0.28, 0.31,0.14, 0.21,
0.43, and0.28,
0.48 m0.31,
3 /m0.43,
3 ; theand 0.48 m
prepared
3/m3; the prepared moisture contents of soil
moisture contents of soil samples from Henan
sampleswere
Province from0.00,
Henan 0.10,Province
0.19, 0.26,were
0.39,0.00,
and 0.10,
0.52 m 0.19,
3 /m0.26,
3 ; and0.39, and 0.52 moisture
the prepared m3/m3; and the prepared
contents of soil
moisture contents of soil samples from Heilongjiang Province were 0.00,
samples from Heilongjiang Province were 0.00, 0.10, 0.17, 0.27, 0.32, 0.44, and 0.56 m /m . The sensor 0.10, 0.17,
3 0.27,
3 0.32, 0.44,
andthen
was 0.56inserted
m /m . into
3 3 The thesensor was
plastic then inserted
tubular column in into
the the
centerplastic tubular column
of measuring barrel toinmeasure
the center eachof
measuring barrel to measure each soil sample ten times. When measurements
soil sample ten times. When measurements were completed, the actual volumetric moisture content of were completed, the
actual
soil samplesvolumetric
under testmoisture content using
was obtained of soilthe samples under method.
oven-drying test was obtained using the oven-drying
method.
3.1.2. Field Test
3.1.2. Field Test
The test was carried out in the Rugao Demonstration Base of National Engineering and Technology
CenterThe for test was carried
Information out in theNanjing
Agriculture, Rugao Demonstration Base of National
Agricultural University from JulyEngineering
to Octoberand 2017. Technology
The soil
Center
type of the for Information
test site was siltyAgriculture,
loam. Field testNanjing
setup Agricultural
and comparison University fromare
between tests July to October
shown in Figure2017.
14.
The soil type of the test site was silty loam. Field test setup and comparison
With a measurement accuracy of 1%, the Diviner2000 portable soil moisture monitoring system was between tests are shown
in Figure
used 14. With a measurement
for comparison. A 120 cm-long accuracy
monitoring of 1%,annular
the Diviner2000
pipe withportable
a 15 cm soil moisture monitoring
aboveground part was
system
and 105 cm wasunderground
used for comparison. A 120 cm-long
part was installed monitoring
to Diviner2000 andannular pipe with
kept sealed in theatest15 cm
plotaboveground
before test,
sopart
thatwastheand 105 cm
system wasunderground
able to measure partthewassoilinstalled
volumetricto Diviner2000 and kept
moisture content ofsealed
verticalin soil
the test plot
profile
before test, so that the system was able to measure the soil volumetric
at the depths of 10, 20, 30, 40, 50, 60, 70, 80, 90, and 100 cm quickly and automatically. Similar to moisture content of vertical
soil profile at
Diviner2000, the depths
a 100 cm-longofmonitoring
10, 20, 30, 40, 50, 60,pipe
annular 70, whose
80, 90, aboveground
and 100 cm quickly and automatically.
and underground parts
Similar to Diviner2000, a 100 cm-long monitoring annular pipe
were both 50 cm was installed to the proposed sensor in the test plot before test, to allow the rapid whose aboveground and
underground
automatic parts were
measurement of both 50 cm
the soil was installed
volumetric moisture to the proposed
content sensor
of vertical in profile
soil the testatplot
the before
depthstest,
of
to allow the
20, 30, and 50 cm. rapid automatic measurement of the soil volumetric moisture content of vertical soil
profile at the depths of 20, 30, and 50 cm.
Sensors 2018, 18, 1648 13 of 23
Sensors 2018, 18, x 13 of 24

Diviner2000 This sensor

Figure14.
Figure 14. Photograph
Photographof
ofthe
thefield
fieldtest
testdesign.
design.

Volumetric moisture contents of different soil layers were obtained during the field test. Then,
Volumetric moisture contents of different soil layers were obtained during the field test. Then,
linear regression analysis of measurement results was carried out. The volumetric moisture content
linear regression analysis of measurement results was carried out. The volumetric moisture content of
of soil layers between 0–100 cm was calculated using the multiple regression equation developed
soil layers between 0–100 cm was calculated using the multiple regression equation developed based
based on sensitive soil layers. The feasibility of moisture content detection based on sensitive layers
on sensitive soil layers. The feasibility of moisture content detection based on sensitive layers was
was verified from the quality of fit between the calculated and measured data.
verified from the quality of fit between the calculated and measured data.
3.1.3. Data Analysis Method
3.1.3. Data Analysis Method
(1) Drying method
(1) Drying method
Samples are classified via the soil volume moisture content volume when the water content of
Samples are classified via the soil volume moisture content volume when the water content of
mass is calculated for the quantity of the samples. The conversion relation between two factors is
mass is calculated for the quantity of the samples. The conversion relation between two factors is
shown as follows:
shown as follows:
θθv v==θθmm××ρρb b (3)
(3)
In
In the
the above equation, θθm represents the mass water content of soil, θv represents the volume
above equation, m represents the mass water content of soil, θ v represents the volume
water content while ρb represents the volume weight of the dry soil.
water
ρbcontent while ρasb the
was calculated represents the volume
ratio between weight
the mass of theofdry
thesoil
dryand
soil.the volume of the dry soil with
ρ b ofwas
the unit g/cm 3 , as followed.
calculated as the ratio between the mass of the dry soil and the volume of the dry soil
with the unit of g/cm , as followed.
3 Ws
ρb = (4)
V0
V0 represents the total volume of the soil ρ
W
= W
while s
s represents the mass of the earth in the soil.
(4)
b
V
The equation for calculating the mass water content of0 soil θm is shown as followed:

V0 represents the total volume of the soil while

Ww W s represents the mass of the earth in the soil.
θm = × 100% (5)
Ws of soil
The equation for calculating the mass water content θm is shown as followed:
In the equation of (5), Ww represents the massWwof water in the soil. The weighing method of oven
drying is applied. The water content value θ = mass
ofmthe × 100 %soil is calculated before the actual volume
of the (5)
Ws
water content of the soil samples is finally obtained from the equation (3).
In the equation
(2) Automatic of (5),
calibration W w represents the mass of water in the soil. The weighing method of
calculation
ovenThe
drying is applied.
sensor Thethe
measures water contentofvalue
frequency of the
the soil massthe
under of the soil isofcalculated
surface the solid before
phase the actual
medium,
volume water
which is X content of the soil samples is finally obtained from the equation (3).
, the gas-phase medium exposed to the air on the surface of the earth of X
iSoil air
and the frequency of liquid-phase medium water wrapped in a hollow water column of Xwater .
(2) Automatic calibration calculation
The sensor measures the frequency of the soil under the surface of the solid phase medium,
which is X iSoil , the gas-phase medium exposed to the air on the surface of the earth of X air and
Sensors 2018, 18, 1648 14 of 23

The self-calibration formula for the three-phase Yiwater normalized frequency value in the soil at the
depth i is:
Xair − X isoil
Yiwater = (6)
Xair − Xwater
Through the relationship between capacitance and dielectric constant ε = Csoil /Cair , the dielectric
constant of the perceived soil can be calculated. Cair is the capacitor when the medium is the air, Csoil is
the capacitor when the medium is soil. Based on the linear relationship between soil permittivity and
√
soil moisture content: θiwater = a ε + b. Among them, a and b are constants and are closely related to
the type of soil to be measured.
(3) Calibration
A soil volumetric moisture content monitoring model was built based on the calibration test.
The root mean square error (RMSE) was used as the index of accuracy of the model:
v
un
u ∑ θvi − θ̂vi
u

t1
RMSE = (7)
n

where θvi is the measured volumetric moisture content of the ith soil sample and θ̂vi is the volumetric
moisture content obtained by the ith soil sample’s sensor.
(4) Consistency Test and Verification Test with Different Soil Types
The sensor’s output stability was tested using the relative error (RE) between the measured value
and the actual value from the consistency test and verification test:
 
θv − θ̂v 
RE = × 100% (8)
θv

where θv is the measured volumetric moisture content of the soil sample and θ̂v is the volumetric
moisture content obtained from the soil sample sensor.
(5) Field Test Accuracy
The field test accuracy of the sensor was determined by comparing the proposed sensor and the
Diviner2000 portable soil moisture monitoring system. The coefficient of determination (R2 ) was used
to evaluate the accuracy:
n  
∑ (y ai − y a ) y pi − y p
R2 = si=1 (9)
n  2
2
∑ (y ai − y a ) y pi − y p
i =1

where y ai and y pi are the measured and predicted moisture content of the whole soil profile at time i,
respectively; y a and y p are the means of n measured and predicted values, respectively. The closer the
R2 to 1, the higher the correlation between the measured and predicted values.
The moisture content of the whole soil profile was predicted using the optimal combination
when the volumetric moisture content measured in sensitive soil layers was the independent variable.
Specifically, soil profile moisture content was verified by means of quantitative inversion. The goodness
of fit between the predicted value and measured data was calculated using the multiple regression
equation built based on sensitive soil layers, through which the feasibility of moisture content detection
based on sensitive soil layers was verified. The volumetric moisture content of soil in deep sensitive
layers served as a controlled variable Xi to calculate the ternary linear regression equation of Y,
the volumetric moisture content of the whole soil profile.
The goodness of fit between the predicted value and measured data was calculated using the
multiple regression equation built based on sensitive soil layers, through which the feasibility of
moisture content detection based on sensitive soil layers was verified. The volumetric moisture
content of soil in deep sensitive layers served as a controlled variable Xi to calculate the ternary
linear regression
Sensors 2018, 18, 1648 equation of Y, the volumetric moisture content of the whole soil profile. 15 of 23
Suppose the linear regression model between random y and general variables X 1 , X 2 ,⋅ ⋅ ⋅ X k
was
Suppose the linear regression model between random y and general variables X1 , X2 , . . . , Xk was
y = β 0 + β1 X 1 + β 2 X 2 + β k X k + ε (10)
y = β 0 + β 1 X1 + β 2 X2 + β k X k + ε (10)
where β1 , β 2 ,⋅ ⋅ ⋅β k are k + 1 unknown parameters; β 0 is regression constant; β1 , β 2 ,⋅ ⋅ ⋅β k is
where β 1 , β 2 , . . . , β k are k + 1 unknown parameters; β 0 is regression constant; β 1 , β 2 , . . . , β k is regression
regression coefficient; y is the explained variable; and X 1 , X 2 ,⋅ ⋅ ⋅ X k are general variables that can
coefficient; y is the explained variable; and X1 , X2 , . . . , Xk are general variables that can be controlled
be controlled
precisely, known precisely, known as
as explanatory explanatory
variables. variables. A
A performance performance
comparison andcomparison
verificationand model verification
is shown
model is shown
in Figure 15. in Figure 15.

Figure 15. Process of multiple linear regression verification.

Figure 15. Process of multiple linear regression verification.

As shown above, soil volumetric moisture content measured by Diviner2000 portable soil
As shown
moisture above,system
monitoring soil volumetric moisture
at the depths content
of 20, measured
30, and 50 cm were by Diviner2000 portablevariables
taken as controlled soil moisture {X1,
monitoring system at the depths of 20, 30, and 50 cm were taken as controlled
X2, X3}, which were then used to construct the ternary linear regression equation with {Y variables {X101, ,YX402,, Y
X60
3 },
,
Y70, Y80, Y90, Y100} representing the soil volumetric moisture content measured at the depths of 10, 40,,
which were then used to construct the ternary linear regression equation with {Y 10 , Y 40 , Y 60 , Y 70 , Y 80
Y90 60,
50, , Y100
70,} representing
80, 90, and 100 thecm.
soil Soil
volumetric moisture
volumetric content
moisture measured
content at theby
measured depths of 10, 40, sensor
the proposed 50, 60, 70, at
80, 90, and 100 cm. Soil volumetric moisture content measured by the proposed sensor
the depths of 20, 30, and 50 cm served as the input and the ternary linear regression equation as the at the depths of
20, 30, andmodel
inversion 50 cm toserved as the
calculate theinput and the moisture
volumetric ternary linear regression
content equation
of the other as the inversion
soil layers at the depths model of
to calculate the volumetric moisture content of the other soil layers at the depths
10, 40, 50, 60, 70, 80, 90, and 100 cm, i.e., values of {y10, y40, y60, y70, y80, y90, y100}. of 10, 40, 50, 60, 70,
80, 90, and 100 cm, i.e., values of {y10 , y40 , y60 , y70 , y80 , y90 , y100 }.

3.2. Results

3.2.1. Calibration of the Soil Volumetric Moisture Content Model

As shown in Figure 16, the measurement frequency of the sensor gradually declined as the
volumetric moisture content of soil sample increased. Using the least squares method, the R2 between
the two was found to be 0.9869, indicating that the frequency measured by the sensor was significantly
correlated with the volumetric moisture content of the soil sample under test.
 3.2. Results
3.2.1. Calibration of the Soil Volumetric Moisture Content Model
3.2.1.As
Calibration
shown inofFigure
the Soil
16,Volumetric Moisturefrequency
the measurement Content Model
of the sensor gradually declined as the
volumetric
As shown moisture content
in Figure of soil
16, the sample increased.
measurement frequencyUsing
of thethe leastgradually
sensor squares method,
declined the R2
as the
between
Sensors the1648
2018, 18,
volumetric two was found
moisture contentto of
be soil
0.9869, indicating
sample that the
increased. frequency
Using measured
the least squaresbymethod,
the sensor
thewas
16Rof2 23
significantly correlated with the volumetric moisture content of the soil sample under test.
between the two was found to be 0.9869, indicating that the frequency measured by the sensor was
significantly correlated with the
8.6
volumetric moisture content of the soil sample under test.

）kHz）
y = 2.2779x2 - 3.2722x + 8.5301
8.6
8.4 R2 = 0.9869

（kHz（
y = 2.2779x2 - 3.2722x + 8.5301

frequency
8.4
8.2 R2 = 0.9869

frequency
8.2
8.0

The measurement
8.0
7.8

The measurement
7.8
7.6 The measurement frequency

The measurement frequency7.6

7.4
0.1 0.2 0.3 0.4 0.5 0.0
7.4 Soil volumetric moisture content(m3/m3)
0.0 0.1 0.2 0.3 0.4 0.5
Figure 16. Calibration
Soil result of soil
volumetric samples
moisture from
content(m3
/m3)Rugao (loam).
Figure 16. Calibration result of soil samples from Rugao (loam).
Figure 16. Calibration result of soil samples from Rugao (loam).
Due to the error exists between the actual value of the system frequency and the nominal value,
it Due
will to
Due
the errordeviation
cause
to thethe
exists between
between
error exists between
thethe
actual valuefrequency
testing
the actual
of the system
value of the system
frequency
output from the
frequency
and the nominal
and sensor
value,
and actual
the nominal value,
it will cause the
frequency. To deviation
gain the between the
conversion testing frequency
relationship between output
the fromtesting
actual the sensor and actual
frequency and frequency.
the soil
it will cause the deviation between the testing frequency output from the sensor and actual
To volume
gain themoisture
conversion relationship between the actual testing frequency theand the soil volume moisture
frequency. To gaincontent, the testing
the conversion frequency
relationship is normalized
between with testing
the actual establishment
frequency andof function
the soil
relationship
content,
volume between
themoisture
testing normalized
frequency
content, resultsfrequency
is testing
the normalizedand with
the soil
is volume
the moisture
establishment
normalized content.
withof As relationship
function
the shown inofFigure
establishment 17,
between
function
the R2 of this
normalized relationship
results and thewas 0.9981.
soil volume moisture content. As shown in Figure 17, the R 2 of this
relationship between normalized results and the soil volume moisture content. As shown in Figure 17,
relationship was
the R2 of this 0.9981. was 0.9981.
relationship
/m3) /m )

0.5
3

y = 0.9129x2 - 0.1188x - 0.0072

3
content(m

0.5
0.4 R2 = 0.9981
y = 0.9129x2 - 0.1188x - 0.0072
3
content(m

0.4
0.3 R2 = 0.9981
moisture

0.3
0.2
moisture
Soil volumetric

0.2
0.1
Soil volumetric

0.1
0.0 Soil volumetric moisture content

0.1 0.2 0.3 0.4volumetric

Soil 0.5 moisture
0.6 0.7
content 0.8
0.0
Normalization frequency(NV)
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Figure 17. Normalization
Normalized frequency(NV)
calibration result.

Figure 17. Normalized calibration result.

Soil volumetric moisture Figure content Normalizedmodel
17.monitoring calibration
builtresult.
based on the calibration test was
verified. As shown in Figure 18, the R 2 was 0.9991 and RMSE
Soil volumetric moisture content monitoring model built based on was 0.0186, indicating good accuracy
the calibration of
test was
the proposed sensor.
Soil volumetric
verified. The
As shownmoisture objective
in Figurecontent of the nominal
18, the Rmonitoring test
2 was 0.9991model is
and RMSEto determine
built was
based the
on the
0.0186, direct relationship
calibration
indicating between
testaccuracy
good was verified.
of
Asthe
the
collection
shown
proposed
circuit
in Figure
sensor.
output
18, R2frequency
theobjective
The wasof0.9991andand
the RMSE
the nominal
water content
test iswas
of actual
0.0186,
to determine
soil value.
indicating goodThe measuring
accuracy
the direct relationship of the
between
algorithm
proposed of the
sensor.
the collection sub-modules
The
circuit objective for
of converting
the nominal
output frequency soiltest
and the water
is to
water content in actual
determine
content of thethe
software
direct system is measuring
confirmed
relationship
soil value. The between
via the nominal
thealgorithm test
collectionofcircuit results.
output frequency
the sub-modules and the
for converting soilwater
watercontent
content of actual
in the soil value.
software systemThe measuring
is confirmed
via the nominal
algorithm test results. for converting soil water content in the software system is confirmed via
of the sub-modules
the nominal test 18,
Sensors 2018, results.
x 17 of 24

0.5
y=0.997x-0.0081
Measured Values（m3/m3）

0.4 R2=0.9963
RMSE=0.019

0.3

0.2

0.1
Jiangsu

0.0
0.0 0.1 0.2 0.3 0.4 0.5
Actual Values（m3\m3）

Figure
Figure 18.18.Verification
Verificationresult
resultofofsoil
soilvolumetric
volumetric moisture
moisture content
contentmonitoring
monitoringmodel.
model.

3.2.2. Results of the Consistency Test

The results of consistency test are shown in Table 2. With the frequency measured by
water-layer probe as the reference value, the relative errors between frequency values measured by
the four other probes and reference value were calculated to be in the range 0–1.17%, indicating that
the results measured by five probes showed favorable consistency.
 Jiangsu

0.0
0.0 0.1 0.2 0.3 0.4 0.5
Actual Values（m3\m3）

Figure 18. Verification result of soil volumetric moisture content monitoring model.
Sensors 2018, 18, 1648 17 of 23
3.2.2. Results of the Consistency Test
TheResults
3.2.2. resultsof of
theconsistency
Consistencytest
Test are shown in Table 2. With the frequency measured by
water-layer probe as the reference value, the relative errors between frequency values measured by
The results of consistency test are shown in Table 2. With the frequency measured by water-layer
the four other probes and reference value were calculated to be in the range 0–1.17%, indicating that
probe as the reference value, the relative errors between frequency values measured by the four other
the results measured by five probes showed favorable consistency.
probes and reference value were calculated to be in the range 0–1.17%, indicating that the results
measured by five Tableprobes showed favorable
2. Measurement consistency.
results and relative error (RE) from the five probes.

Volumetric Table 2. Measurement resultsAir and relative error (RE) from the five probes.
Water
50 cm 30 cm 20 cm 50 cm 30 cm 20 cm Air layer
Moisture Layer Layer
(kHz) (kHz) (kHz) RE (%) RE (%) RE (%) RE (%)
Content (m /m )Moisture
3
Volumetric3
50 cm 30 cm 20 cm (kHz)
Air Layer (kHz)
Water Layer 50 cm 30 cm 20 cm Air layer
Content (m3 /m3 ) (kHz) (kHz) (kHz) (kHz) (kHz) RE (%) RE (%) RE (%) RE (%)
0.00 8.57 8.56 8.58 8.56 8.55 0.23 0.12 0.35 0.12
0.07 0.00 8.33 8.57
8.31 8.568.34 8.58 8.33 8.56 8.35 8.55 0.24 0.23 0.48 0.12 0.120.35 0.12
0.24
0.07 8.33 8.31 8.34 8.33 8.35 0.24 0.48 0.12 0.24
0.13 8.06 8.18 8.16 8.09 8.11 0.62 0.86 0.62 0.25
0.13 8.06 8.18 8.16 8.09 8.11 0.62 0.86 0.62 0.25
0.20 0.20 7.99 8.00
7.99 8.007.98 7.98 7.97 7.97 7.99 7.99 0.00 0.00 −0.13−0.13 0.130.13 0.25
0.25
0.26 0.26 7.85 7.84
7.85 7.847.82 7.82 7.86 7.86 7.83 7.83 0.26 0.26 −0.13−0.13 0.130.13 0.38
0.38
0.31 0.31 7.76 7.75
7.76 7.757.75 7.75 7.76 7.76 7.75 7.75 0.13 0.13 0.00 0.00 0.000.00 0.13
0.13
0.37 0.37 7.68 7.68
7.65 7.657.62 7.62 7.72 7.72 7.63 7.63 0.66 0.66 0.26 0.26 0.130.13 1.18
1.18
0.47 0.47 7.51 7.51
7.52 7.527.51 7.51 7.51 7.51 7.52 7.52 0.13 0.13 0.00 0.00 0.130.13 0.13
0.13
0.59 0.59 7.38 7.38
7.36 7.367.35 7.35 7.37 7.37 7.39 7.39 0.14 0.14 0.41 0.41 0.540.54 0.27
0.27

Figure
Figure1919shows the regression
shows the regressionbetween
between measurement
measurement frequencies
frequencies ofprobes
of five five probes
and theand the
measured
measured soil moisture
soil moisture contents. contents.
The fittingThe fitting
curves curves
of five of five
probes wereprobes were overlapping
overlapping in the figure, in the figure,
indicating good
indicating good consistency between the sensor probes developed
consistency between the sensor probes developed in this study. in this study.

8.5 20cm
The frequency measured(kHz)

30cm
50cm
Air
8.0 Water

7.5

0.0 0.2 0.4 0.6

Soil volumetric moisture content(m3/m3)

Figure 19. Fitting result between the values measured by five probes and soil volumetric moisture
Figure 19. Fitting result between the values measured by five probes and soil volumetric
content.
moisture content.

3.2.3. Verification of Different Soil Types

Regression analysis between volumetric moisture content of different types of soil samples and
soil moisture content measured by the sensor (Figure 20) indicated that the R2 and RMSE of soil from
Jiangxi Province (sandy loam) were 0.9962 and 0.0535, respectively; those of soil from Henan Province
(sandy loam) were 0.9774 and 0.0300, respectively; and those of soil from Heilongjiang Province (clay
loam) were 0.9623 and 0.0364, respectively. As a whole, the R2 of linear regression between measured
values and actual values in all four regions was 0.9644 and the RMSE was 0.0423. The R2 values
between the soil moisture measured by the proposed sensor and actual soil moisture content were
higher than 0.96 for all soil types. They were significantly correlated, with a small RMSE, suggesting
that the proposed sensor was favorably adaptable to the measurement of volumetric moisture content
of various soil types.
 Jiangxi Province (sandy loam) were 0.9962 and 0.0535, respectively; those of soil from Henan
Province (sandy loam) were 0.9774 and 0.0300, respectively; and those of soil from Heilongjiang
Province (clay loam) were 0.9623 and 0.0364, respectively. As a whole, the R2 of linear regression
between measured values and actual values in all four regions was 0.9644 and the RMSE was 0.0423.
The R2 values between the soil moisture measured by the proposed sensor and actual soil moisture
content were higher than 0.96 for all soil types. They were significantly correlated, with a small
Sensors 2018, 18, 1648 18 of 23
RMSE, suggesting that the proposed sensor was favorably adaptable to the measurement of
volumetric moisture content of various soil types.

0.6 0.6
y=1.192x-0.0071 y=1.077x-0.0091

Measured Values（m3/m3）

Measured Values（m3/m3）
0.5 R2=0.9884 0.5 R2=0.9774
RMSE=0.054 RMSE=0.030
0.4 0.4

0.3 0.3

0.2 0.2

0.1 Jiangxi 0.1 Henan

0.0 0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Actual Values（m3/m3） Actual Values（m3/m3）
(Jiangxi) (Henan)
0.6
y=1.1203x-0.0406 0.6 y=1.0531+0.002
Measured Values（m3/m3）

0.5

Measured Values（m /m ）
R2=0.9623 R2=0.9711

3
RMSE=0.036

3
0.4
0.4
0.3

0.2 0.2 Henan

Heilongjiang
0.1 Heilongjiang Jiangxi
0.0 Jiangsu
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.0 0.2 0.4 0.6
Actual Values（m3/m3） Actual Values（m3/m3）
(Heilongjiang) (Integrated)

Figure 20. Integrated verification of volumetric moisture content of different types of soil samples.
Figure 20. Integrated verification of volumetric moisture content of different types of soil samples.
3.2.4. Field Test Accuracy
3.2.4. Field TestComparison
Accuracyof measurement results between the proposed sensor and the Diviner2000 soil
moisture monitoring system is shown in Figure 21. The R2 and RMSE between the value measured
Comparison of measurement
by the proposed sensor and soilresults between
volumetric moisturethe proposed
content sensor were
from Diviner2000 and0.86theand
Diviner2000
1.75, soil
respectively, at 20 cm soil depth; 0.87 and 0.87, 2
respectively, at 30 cm soil
moisture monitoring system is shown in Figure 21. The R and RMSE between the value measured depth; 0.86 and 0.32,
respectively, at 40 cm soil depth; and 0.94 and 0.38, respectively, at 50 cm soil depth.
by the proposed sensor and soil volumetric moisture content from Diviner2000 were 0.86 and 1.75,
respectively, at 20 cm soil depth; 0.87 and 0.87, respectively, at 30 cm soil depth; 0.86 and 0.32,
respectively, at 40 cm soil depth; and 0.94 and 0.38, respectively, at 50 cm soil depth.
Sensors 2018, 18, x 19 of 24

47
39 y = 0.8029x + 6.2268 y = 0.7114x + 13.429
R2 = 0.86
Proposed sensor（m3/m3）

Proposed sensor （m3/m3）

R2 = 0.87
38 RMSE=1.75 46 RMSE=0.88

37
45
36
44
35
20cm 30cm
34 43
34 35 36 37 38 39 40 41 42 43 44 45 46
Diviner2000（m3/m3） Diviner2000（m3/m3）
(a) 20 cm (b) 30 cm
50 7
y = 0.767x + 11.456 20cm 30cm 50cm
6
Proposed sensor （m3/m3）

R2 = 0.86
49
relative error （% ）

RMSE=0.32 5

4
48 3

2
47
1
50cm
0
46
46 47 48 49 0 5 10 15 20 25 30 35
Diviner2000（m3/m3） Observation number
(c) 50 cm (d) relative error

Figure 21. Verification of measuring performance at (a) 20 cm, (b) 30 cm, (c) 50 cm, and (d) relative error.
Figure 21. Verification of measuring performance at (a) 20 cm, (b) 30 cm, (c) 50 cm, and (d) relative error.
According to the relative errors between the results measured by the proposed sensor and
Diviner2000 (Figure 21d), the proposed sensor’s maximum relative measurement error was 5.73%
and it was smaller than 5% at for all soil depths except 20 cm. This indicates that the measurement
accuracy of the proposed sensor was comparable with that of Diviner2000 and met the requirements
for application.
Ternary linear regression of soil volumetric moisture content at depths of 10, 40, 50, 60, 70, 80,
90, and 100 cm were constructed with the soil volumetric moisture content measured at the depths of
Sensors 2018, 18, 1648 19 of 23

According to the relative errors between the results measured by the proposed sensor and
Diviner2000 (Figure 21d), the proposed sensor’s maximum relative measurement error was 5.73%
and it was smaller than 5% at for all soil depths except 20 cm. This indicates that the measurement
accuracy of the proposed sensor was comparable with that of Diviner2000 and met the requirements
for application.
Ternary linear regression of soil volumetric moisture content at depths of 10, 40, 50, 60, 70, 80,
90, and 100 cm were constructed with the soil volumetric moisture content measured at the depths of
20, 30, and 50 cm by Diviner2000 as the controlled variable. As shown in Table 3, the R2 of regression
equations for predicted depths of 10 cm, 40 cm, and 60 cm are higher than 0.9, and the R2 of regression
equations for other predicted depths are higher than 0.85. The highest R2 of 0.984 was found for
regression equation for predicted depths of 40 cm layer in a loam soil.

Table 3. Fit of the ternary regression equation.

Predicted Depth (cm) Ternary Regression Equation R2 RMSE

10 y10 = 1.403X1 −0.065X2 − 0.313X3 0.915 2.1
40 y40 = 0.057X1 + 0.875X2 + 0.112X3 0.984 0.72
60 y60 = 0.093X1 + 0.008X2 + 0.684X3 0.976 0.27
70 y70 = 0.195X1 + 0.396X2 − 0.065X3 0.862 0.31
80 y80 = 0.187X1 + 0.413X2 − 0.0673X3 0.861 0.32
90 y90 = 0.196X1 − 0.413X2 − 0.068X3 0.865 0.32
100 y100 = 0.196X1 + 0.417X2 − 0.069X3 0.863 0.33

The volumetric
Sensors 2018, 18, x moisture content of 10 cm soil depth measured by Diviner2000 was taken20asofthe 24
measured value, and ternary regression was carried out with the soil volumetric moisture contents
measured by the the proposed
proposedsensor
sensoratatthe
thedepths
depthsofof20,20,
30,30,
andand
50 50
cmcm (Figure
(Figure 22).22).
TheThe R2 was
R2 was 0.84,0.84,
and
and the relative
the relative errorerror between
between fitted
fitted values
values and and measured
measured value
value waswas 0.06–1.48%
0.06–1.48% (<5%).
(<5%).

1.6
y = 1.1173x - 0.9783 1.48
30 R2 = 0.84
Measured value（m3/m3）

RMSE=2.1 1.2
Relative error(%)

28

0.8
26

24 0.4

22 0.06
22 24 26 28 0.0
0 5 10 15 20 25 30 35
Fitted value （m3/m3） Observation number

(a) 10 cm (b) relative error

Figure 22. Correlation

22. Correlation and
and relative
relative error
error between
between fittedfitted
valuevalue and measured
and measured value
value at at theofdepth
the depth 10 cm.of
10 cm.
Fitting of volumetric moisture content of the other soil layers was carried out according to the
volumetric
Fitting moisture content
of volumetric measured
moisture by the
content proposed
of the sensor
other soil layersat was
the depths
carriedofout20,according
30, and 50 to cm.
the
The correlation between the fitting values of 40, 60, 70, 80, 90, and 100 cm-deep
volumetric moisture content measured by the proposed sensor at the depths of 20, 30, and 50 cm. soil layers and
measured
The valuesbetween
correlation (Figure 23)
theproduced R2 values
fitting values of 40,of60,
more70,than 0.8, and
80, 90, which100indicated
cm-deepthat soilthe moisture
layers and
content of the other soil layers could be2 calculated based on the moisture content
measured values (Figure 23) produced R values of more than 0.8, which indicated that the moisture of sensitive soil
layers measured by the proposed sensor. The soil volumetric moisture content of soil layers
content of the other soil layers could be calculated based on the moisture content of sensitive soil layers between
0–100 cm by
measured was
the favorably inverted
proposed sensor. Theby
soilthe proposed
volumetric sensor,content
moisture providing
of soilalayers
new between
approach to the
0–100 cm
development of an efficient soil moisture sensing system.

50
ue of 70cm（m3/m3）
ue of 60cm（m3/m3）

44
e of 40cm（m3/m3）

y = 1.223x - 10.057 y = 1.1338x - 6.5342 y = 1.1746x - 8.3517

R2 = 0.88 49
R2 = 0.89 R2 = 0.82
43 RMSE=0.72 RMSE=0.27 RMSE=0.31
49

42
48
 Fitting of volumetric moisture content of the other soil layers was carried out according to the
volumetric moisture content measured by the proposed sensor at the depths of 20, 30, and 50 cm.
The correlation between the fitting values of 40, 60, 70, 80, 90, and 100 cm-deep soil layers and
measured values (Figure 23) produced R2 values of more than 0.8, which indicated that the moisture
Sensors 2018, 18, 1648 20 of 23
content of the other soil layers could be calculated based on the moisture content of sensitive soil
layers measured by the proposed sensor. The soil volumetric moisture content of soil layers between
0–100
was cm wasinverted
favorably favorablyby inverted by the
the proposed proposed
sensor, sensor,a new
providing providing a new
approach approach
to the to the of
development
andevelopment
efficient soilofmoisture
an efficient soil moisture
sensing system. sensing system.

50

The measured value of 70cm（m3/m3）

The measured value of 60cm（m3/m3）
44
The measured value of 40cm（m3/m3）

y = 1.223x - 10.057 y = 1.1338x - 6.5342 y = 1.1746x - 8.3517

R2 = 0.88 49
R2 = 0.89 R2 = 0.82
43 RMSE=0.72 RMSE=0.27 RMSE=0.31
49

42
48
48
41

40 47
47
41 42 43 44 47 48 49 50 47 48 49
The fitting value of 40cm（m3/m3） The fitting value of 60cm（m3/m3） The fitting value of 70cm（m3/m3）

The measured value of 100cm（m3/m3）

51
The measured value of 90cm（m3/m3）
The measured value of 80cm（m3/m3）

y = 1.1727x - 8.5097
y = 1.1756x - 8.4767 y = 1.173x - 8.5767 50 R2 = 0.82
49 R2 = 0.81 R2 = 0.81 RMSE=0.33
RMSE=0.32 RMSE=0.32
50

49
48
49

48
47
48 48 49 50
47 48 49 48 49 50 The fitting value of 100cm（m3/m3）
The fitting value of 80cm（m3/m3） The fitting value of 90cm（m3/m3）

Figure23.
Figure 23. Correlation
Correlation between
betweenfitted
fittedvalues
valuesand
andmeasured
measuredvalues at at
values different depths.
different depths.

4. Discussion
4. Discussion
Development of an accurate soil profile moisture sensor remains the priority for intelligent
Development of an accurate soil profile moisture sensor remains the priority for intelligent
farmland irrigation because traditional needle-type soil moisture sensors are vulnerable to erosion
farmland irrigation because traditional needle-type soil moisture sensors are vulnerable to erosion due
due to direct contact with soil, which results in low measurement accuracy [4]. In addition,
toinadequate
direct contact with
needle soil, which
length results
also makes it in low measurement
difficult accuracy
to conduct simple and[4]. In addition,
quick moisture inadequate
sensing
needle length also makes it difficult to conduct simple and quick moisture sensing content in deep soil.
content in deep soil. To this end, a non-contact measurement structure based on annular-electrode
To this end, a non-contact measurement structure based on annular-electrode probes was proposed in
this study. The axial radiation diameter and range of radial plane radiation of the sensor’s sensing
probes were 94.5 mm and 100 mm, respectively. The measurement radius of this sensor was 2.5 times
that of the equally spaced three-depth soil profile moisture sensor developed by Yuan et al. [19].
Unlike TDR [8], the HFC method developed in this study can be directly applied to soil profile
moisture monitoring. A sensor probe control circuit that enabled moisture content at different soil
depths by means of time-sharing power supply was designed to avoid signal interference and power
overload. The AVR single-chip microcontroller served as the MCU. Functions such as measuring the
output frequency of acquisition circuit were realized at the hardware level, which enables the sensor to
meet practical measurement demands.
According to the comparison between the proposed sensor and Diviner2000 portable soil moisture
monitoring system, relative error between the values measured by both sensors at the depth of 20 cm
was higher than those at the depths of 30 cm and 50 cm, which might be because the 20 cm soil layer is
close to the soil surface and the temperature of the topsoil is higher than that of the deep soil from July
to September when the test was conducted. The field test of sensor accuracy indicated that quantitative
inversion of moisture content of the other soil layers at the depth of 0–100 cm could be realized based
on the moisture content of three soil layers measured by the proposed sensor. The R2 of regression
equation between measured values and fitted values was higher than 0.85, which demonstrated
favorable inversion of the volumetric moisture content at 0–100 cm soil depth by the proposed sensor.
The findings of this research provide a new direction for the development of efficient moisture
sensing systems. However, the proposed sensor might not be accurate in arid land since the modeling
data were collected from the rice field. Therefore, further modeling is still required to verify its inversion
function under other environmental conditions. Because only the influence of moisture variations
in different soil types on sensor output was the focus of this study, the influence of physicochemical
Sensors 2018, 18, 1648 21 of 23

properties (e.g., soil temperature, volumetric weight, and conductivity) on the sensor’s measurement
should be considered in future research.

5. Conclusion
A soil profile moisture sensor was developed based on the principle of HFC in this study.
The simulation of sensor probe structure, design software, and hardware systems, and tests on
the proposed sensor were also conducted.

(1) Soil moisture measurement following the HFC principle was used to design a sensor that can meet
the demands for soil profile moisture measurement. A sensor that is able to conduct real-time
detection of volumetric moisture content in three sensitive soil layers was proposed. Double
copper rings were used as the component of annular electrodes of the proposed sensor’s sensing
probe. Different structure sizes and electrode spacing of annular electrodes were simulated using
HFSS. The probe structure for an optimal detection range was determined by means of simulation
analysis (W = 30 mm and D = 40 mm).
(2) The hardware circuit was designed using the high-frequency LC in parallel with a resonance
circuit comprising the voltage-controlled oscillator and annular electrodes. Frequency division
and conditioning of frequency signals were performed. A software system was designed to realize
real-time detection of soil profile moisture content. The calibration test of soil volumetric moisture
content model produced an R2 of 0.9663. According to the performance test, different sensor
probes had good consistency, with absolute relative errors between 0% and 1.17%. The sensor
shows favorable adaptability to different soil types. Fitting between actual volumetric moisture
content and values measured by the proposed sensor showed an R2 and RMSE of 0.9644 and
0.0423, respectively.
(3) According to the results of the sensor accuracy test, the R2 values between results measured by
the proposed sensor and Diviner2000 were above 0.85, with relative errors less than 5%. With the
soil volumetric moisture content measured at the depths of 20, 30, and 50 cm measured by the
proposed sensor as the controlled variable, the R2 values between the calculated and measured
soil moisture values of the other soil layers were all larger than 0.80 and thus quantitative
inversion of volumetric moisture content of the other soil layers at the depth of 0–100 cm was
successful. In conclusion, the sensor designed in this study has shown a promising level of
performance and can be applied to practical measurement of soil profile moisture content at
different depths.
(4) The test and the inversion of the water content in the soil profile are realized well from the sensing
methods and the sensor developed here. However, current tested data are derived only from
the soil in four different areas. Meanwhile, only the samples from the rice fields in plain areas
are taken based on the analysis results of the sensitive soil layers. Before applying our sensor in
the dry farms and other hilly areas on a large scale, compatibility tests should be carried out to
obtain accurate and stable results from the sensor.

Author Contributions: Y.Z., W.C., and J.N. conceived and designed the experiments; G.Z., Z.Q., and C.L.,
performed the experiments; Z.G., and Z.Q. analyzed the data; Z.G., Y.Z., and J.N. wrote the paper.
Funding: This work was supported by grants from the National Key Research and Development Program of
China (2016YFD0300607), the Three New Project of Agricultural Machinery in Jiangsu Province (NJ2017-23),
the Independent Innovation of Agricultural Science and Technology in Jiangsu Province (SCX(16)1006), the Priority
Academic Program Development of Jiangsu Higher Education Institutions (PAPD), and the 111Project (B16026).
Conflicts of Interest: The authors declare no conflict of interest.
Sensors 2018, 18, 1648 22 of 23

References
1. Li, L.; Sun, Y.; Lin, J. A wireless-sensor for soil water content powered by solar energy. J. Jiangsu Univ.-Nat.
Sci. Ed. 2009, 30, 541–544.
2. Coelho, E.F.; Or, D. Flow and uptake patterns affecting soil water sensor placement for drip irrigation
management. Trans. ASAE 1996, 39, 2007–2016. [CrossRef]
3. Aljoumania, B.; Sànchez-Espigares, J.A.; Cañameras, N.; Josaa, R.; Monserratc, J. Time series outlier and
intervention analysis: Irrigation management influenceson soil water content in silty loam soil. Agric. Water Manag.
2012, 111, 105–114. [CrossRef]
4. Jia, Z.; Zhu, H.; Wang, J.; Ji, G. Application of Sensor Technology in Soil Moisture Monitoring based on
Dielectrometry. Chin. Agric. Sci. Bull. 2015, 31, 246–252.
5. Sun, Y.; Ma, D.; He, Q.; Zhao, Y. Experimental study of a sensor for real-time measurement of soil moisture
profile. J. Beijing For. Univ. 2006, 28, 55–59.
6. Wang, K.; Chan, Y. Research on Soil Moisture Measurement Technology and Soil Moisture Monitoring System;
Machinery Industry Press: Beijing, China, 2011; pp. 132–145.
7. Sheng, W.; Sun, Y.; Lammers, P.S.; Schumann, H.; Berg, A.; Shi, C.; Wang, C. Observing soil water dynamics
under two field conditions by a novel sensor system. J. Hydrol. 2011, 409, 555–560. [CrossRef]
8. Li, X.; Wu, Y.; Li, G.; Zeng, Q.; Wang, S. Development of wireless soil moisture sensor based on solar energy.
Trans. Chin. Soc. Agric. Eng. 2010, 26, 13–18.
9. Zhao, Y.; Ma, Y.; Wang, Y. Green land precision irrigation control system and analysis of optimal irrigation
amount. Trans. Chin. Soc. Agric. Mach. 2012, 43, 46–50.
10. Evett, S.R.; Schwartz, R.C.; Casanova, J.J.; Heng, L.K. Soil water sensing for water balance, ET and WUE.
Agric. Water Manag. 2012, 104, 1–9. [CrossRef]
11. He, L.; Ma, Y.; Zhang, J. Basic characteristics study of RV-1 TDR soil moisture sensor. Ground Water 2016, 3,
18–20.
12. Amato, M.; Ritchie, J.T. Small Spatial Scale Soil Water Content Measurement with Time-Domain
Reflectometry. Soil Sci. Soc. Am. J. 1995, 59, 325–329. [CrossRef]
13. Maheshwarla, S.V.; Venkatasubramanian, R.; Boehm, R.F. Comparison of Time Domain Reflectometry
Performance Factors for Several Dielectric Geometries: Theory and Experiments. Water Resour. Res. 1995, 31,
1927–1934. [CrossRef]
14. Jiang, Z.; Tan, C.; Zhi, X.; Wang, C.; Ma, Y. Development of portable soil moisture detector based on principle
of frequency domain reflectometry. Transducer Microsyst. Technol. 2013, 32, 79–82.
15. Hilhorst, M.A.; Breugel, K.V.; Kroesea, W.S. Dielectric Sensors Used in Environmental and Construction
Engineering. MRS Online Proc. Libr. Arch. 1995, 411. [CrossRef]
16. Balendonck, J.; Hilhorst, M.A.; Whalley, W.R. Sensors for Soil, Substrates, and Concrete Based on the MCM100
Microchip; Electromagnetic Aquametry; Springer: Berlin/Heidelberg, Germany, 2005; pp. 277–315.
17. Heimovaara, T.J. Frequency domain analysis of time domain reflectometry waveforms, 1, Measurements of
the complex dielectric permittivity of soils. Water Resour. Res. 1994, 30, 189–199. [CrossRef]
18. Li, J.; Hong, T.; Feng, R. High-frequency capacitive soil water content sensor based on detecting of true root
mean square. Trans. Chin. Soc. Agric. Eng. 2011, 27, 216–221.
19. Yun, Y.; Sheng, W.; Sun, Y. Design and experiment of soil water sensor monitoring soil moisture at three
depths. Trans. Chin. Soc. Agric. Eng. 2014, 30, 64–71.
20. Zhou, H.; Sun, Y.; Schulze Lammers, P. Dielectric Soil Moisture Sensing Technique Based on Horizontal Scale
Expansion. Trans. Chin. Soc. Agric. Mach. 2015, 46, 261–265.
21. Fu, L.; Sun, Y. Design and applicable evaluation of instrument for determining soil moisture profile.
Trans. CSAE 2010, 26, 42–46.
22. Jabro, J.D.; Leib, B.G.; Jabro, A.D. Estimating soil water content using site-specific calibration of capacitance
measurements from Sentek EnviroSCAN systems. Appl. Eng. Agric. 2005, 21, 393–399. [CrossRef]
23. Tang, Z.C. New developments of the moisture detection and control in the press part of papermachine
EnviroScan~(TM) post press moisture detection system. China Pulp Paper Ind. 2004, 6, 003.
24. Babakutigi, A.N.; Dim, L.A.; Odunze, A.C.; Umar, I.M. Soil water balance approach in root zone of maize
(95 TZEE-Y) using capacitance probe (Diviner 2000) in northern Guinea Savannah of Nigeria. Bayero J. Pure
Appl. Sci. 2010, 3, 87–90.
Sensors 2018, 18, 1648 23 of 23

25. Zhan, H.; Wang, H.; Li, C.; Wan, C. The soil moisture sensor based on soil dielectric property.
Pers. Ubiquitous Comput. 2016, 1, 67–74.
26. Zhao, Y.; Wang, Y. Intelligent system of measuring the spatial distributions of soil moisture. Trans. Chin. Soc.
Agric. Mach. 2005, 36, 76–78.
27. Starks, P.J.; Heathman, G.C.; Ahuja, L.R.; Ma, L. Use of limited soil property data and modeling to estimate
root zone soil water content. J. Hydrol. 2003, 272, 131–147. [CrossRef]
28. Zhang, Y.; Ma, Y.; Jiang, Z. Progress on Research and Application of Rapid Measurement of Soil Moisture
Sensor. Chin. Agric. Sci. Bull. 2014, 5, 170–174.
29. Peterson, A.M.; Helgason, W.D.; Ireson, A.M. Estimating field-scale root zone soil moisture using the
cosmic-ray neutron probe. Hydrol. Earth Syst. Sci. Discuss. 2016, 12, 12789–12826. [CrossRef]
30. Wilczek, A.; Szypłowska, A.; Skierucha, W.; Cie(x0015B)la, J.; Pichler, V.; Janik, G. Determination of Soil
Pore Water Salinity Using an FDR Sensor Working at Various Frequencies up to 500 MHz. Sensors 2012, 12,
10890–10905. [CrossRef] [PubMed]
31. Ritter, A.; Regalado, C.M. Corrections for Simultaneous Measurements of Soil Water Content and Salinity
Using a Capacitance Sensor. In Proceedings of the Annual Meeting on American Society of Agricultural and
Biological Engineers (2007 ASAE), Minneapolis, MN, USA, 17–20 June 2007.
32. Scudiero, E.; Berti, A.; Teatini, P.; Morari, F. Simultaneous Monitoring of Soil Water Content and Salinity
with a Low-Cost Capacitance-Resistance Probe. Sensors 2012, 12, 17588–17607. [CrossRef] [PubMed]
33. Zhu, A.N.; Ji, L.Q.; Zhang, J.B.; Xin, X.L.; Liu, J.L. Research progress on soil moisture measurement via
ground-penetrating radar. Chin. J. Eco-Agric. 2009, 17, 1039–1044. [CrossRef]
34. Gao, Z.; Liu, W.; Zhao, Y.; Zhang, C. Design and Performance Analysis of Composite Sensor for Multilayer
Soil Profile. Trans. Chin. Soc. Agric. Mach. 2016, 47, 108–117.
35. El Marazky, M.S.A.; Mohammad, F.S.; Al-Ghobari, H.M. Evaluation of Soil Moisture Sensors under Intelligent
Irrigation Systems for Economical Crops in Arid Regions. Am. J. Agric. Biol. Sci. 2011, 6, 287–300. [CrossRef]
36. Shi, Y.; Li, F.; Sun, K.; Wang, Z.C.; Dong, Y.H. Study on the moisture sensor laying place in watching and
forecasting soil moisture content. J. Laiyang Agric. Coll. 2006, 23, 179–184.
37. Kun, L.I.; Jianshe, L.I.; Suping, W.U.; Cao, Y.; Gao, Y. Research of Soil Moisture Sensor Location Based on the
Real-time Control of Irrigation Systems on Greenhouse Cucumber. J. Irrig. Drain. 2015, 34, 18–23.
38. Gao, Y.; Liu, Z.D.; Duan, A.W.; Liu, Z.G.; Zhang, J.Y.; Sun, J.S. Laying Depth of Soil Moisture Probe in Soil
Profile and Data Processing Technology. J. Irrig. Drain. 2011, 30, 28–32.
39. Wang, D.; Shu, Y. Research Progress in Determination Methods for Soil Water Content. J. Mt. Agric. Biol.
2017, 2, 61–65.
40. Hilhorst, M.A. Dielectric Characterization of Soil. Ph.D. Thesis, Wageningen Agricultural University,
Wageningen, The Netherlands, 3 February 1998.
41. Kelleners, T.J.; Soppe, R.W.O.; Robinson, D.A.; Schaap, M.G.; Ayars, J.E.; Skaggs, T.H. Calibration of
Capacitance Probe Sensors using Electric Circuit Theory. Soil Sci. Soc. Am. J. 2004, 68, 430–439. [CrossRef]
42. He, Q.; He, J. The principle and application of a single planar capacitance sensor. Ind. Instrum. Autom. 2001,
5, 62–63.

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).