sensors

Article
Low-Cost Soil Moisture Profile Probe Using
Thin-Film Capacitors and a Capacitive Touch Sensor
Yuki Kojima 1, *, Ryo Shigeta 2 , Naoya Miyamoto 3 , Yasutomo Shirahama 2 , Kazuhiro Nishioka 3 ,
Masaru Mizoguchi 3 and Yoshihiro Kawahara 2
1 Faculty of Engineering, Gifu University, Gifu 501-1193, Japan
2 Graduate School of Information Science and Technology, The University of Tokyo, Tokyo 113-8656, Japan;
shigeta@akg.t.u-tokyo.ac.jp (R.S.); shirahama@akg.t.u-tokyo.ac.jp (Y.S.); kawahara@akg.t.u-tokyo.ac.jp (Y.K.)
3 Graduate School of Agricultural and Life Sciences, The University of Tokyo, Tokyo 113-8657, Japan;
miyamoto@isas.a.u-tokyo.ac.jp (N.M.); nishioka@isas.a.u-tokyo.ac.jp (K.N.);
amizo@mail.ecc.u-tokyo.ac.jp (M.M.)
* Correspondence: kojima@gifu-u.ac.jp; Tel.: +81-58-293-2418

Academic Editor: Simon X. Yang

Received: 30 May 2016; Accepted: 9 August 2016; Published: 15 August 2016
Abstract: Soil moisture is an important property for agriculture, but currently commercialized soil
moisture sensors are too expensive for many farmers. The objective of this study is to develop a
low-cost soil moisture sensor using capacitors on a film substrate and a capacitive touch integrated
circuit. The performance of the sensor was evaluated in two field experiments: a grape field and a
mizuna greenhouse field. The developed sensor captured dynamic changes in soil moisture at 10,
20, and 30 cm depth, with a period of 10–14 days required after sensor installation for the contact
between capacitors and soil to settle down. The measured soil moisture showed the influence of
individual sensor differences, and the influence masked minor differences of less than 0.05 m3 ·m−3 in
the soil moisture at different locations. However, the developed sensor could detect large differences
of more than 0.05 m3 ·m−3 , as well as the different magnitude of changes, in soil moisture. The price
of the developed sensor was reduced to 300 U.S. dollars and can be reduced even more by further
improvements suggested in this study and by mass production. Therefore, the developed sensor will
be made more affordable to farmers as it requires low financial investment, and it can be utilized for
decision-making in irrigation.

Keywords: soil moisture; soil moisture profile probe; printing circuit; capacitive touch sensor

1. Introduction
Soil moisture is an important property for agriculture, indicative of how much water is available
for plants. Therefore, its measurement can be utilized for decision-making in irrigation [1–3].
Continuous in situ measurement of the soil moisture started around 1940 with the electrical resistance
method using a gypsum block [4]. Electrodes in the block measure the internal electrical resistance,
which is correlated to soil moisture under an equilibrium condition. This method was simple and
cheap, but had problems concerning accuracy, e.g., the electrical resistance of gypsum is affected
by temperature [5], response of the electrical resistance to changes in soil moisture is slow [6], and
relationship between resistance and soil is not universal for a variety of soils [7].
A neutron attenuation probe was developed as a better method to measure in situ soil moisture
in the 1950s [8,9]. The neutron probe measures a count of radioactive beams emitted from a source,
which is correlated to soil moisture. The method was successful in field measurement [10], but a health
hazard problem remained.
The measurement of in situ soil moisture advanced with the discovery of time domain
reflectometry (TDR) [11]. TDR measures apparent permittivity of the soil from the time that an

Sensors 2016, 16, 1292; doi:10.3390/s16081292 www.mdpi.com/journal/sensors

Sensors 2016, 16, 1292 2 of 14

electromagnetic pulse takes to go through parallel electrodes inserted into soil. Because the apparent
permittivity of water (≈80) is much larger than those of other soil constituents (≈1–12), the apparent
permittivity of soil is strongly correlated to soil moisture [12]. The advantages of the TDR method are
that it can measure soil moisture and electrical conductivity simultaneously [13], it is not susceptible
to temperature and salinity [11], and a variety of sampling volumes is possible by changing the design
of the electrodes [14,15]. However, TDR equipment is expensive, e.g., 3000 U.S. dollars or more.
The capacitance soil moisture sensor is known to be a good alternative to TDR [16–19]. Soil
capacitance depends on the apparent permittivity of soil such that capacitance sensors can determine
soil moisture as well as TDR [20]. It has been reported that the capacitance sensors are relatively
sensitive to soil temperature [21], do not work well in some soils [22–24], and have a small sampling
volume [25]. Even with these disadvantages, a variety of capacitance sensors have been developed
and commercialized [1,26,27] because they are less expensive than TDR and sufficiently reliable.
For example, the capacitance sensors commercialized by Decagon Devices Inc. (Pullman, WA, USA)
are known to be relatively inexpensive and reliable, and these sensors have been widely used in recent
scientific research [28–30].
Agriculture in Japan is in transition from traditional farming, which relies on the individual
farmers’ skills and experiences, to precision farming (also called smart farming). Precision farming
utilizes a variety of sensors to monitor and control environmental and crop conditions in order to
achieve stable crop production and reduce the excessive use of resources [31]. For precision farming,
the soil moisture must be measured at multiple locations and at multiple depths. However, with
capacitance sensors, the measurement of soil moisture implies a significant investment. Therefore,
precise management of fields and crops with a number of soil moisture sensors, i.e., a soil moisture
sensor network, is still impeded by the sensor cost.
There is a trade-off between the cost and accuracy of sensors, i.e., more accurate sensors are more
expensive. Most of the currently commercialized soil moisture sensors tend to have good accuracy by
sacrificing cost reduction as they have been developed for research use. For agriculture, price must
have a higher priority than accuracy, i.e., a low-cost sensor even with a relatively weak accuracy is
preferred. The development of low-cost soil moisture sensors has been prevented by the expensive
and complicated circuits for the high-frequency measurement of the capacitance as well as by the
difficulty of obtaining a flexible sensor design that can accommodate a variety of environment and
soil types. However, it is possible for soil moisture sensors to be much less expensive than currently
commercialized ones, owing to recent technological developments. Thus, the objective of this study is
to develop a low-cost soil moisture sensor with the latest technologies, i.e., coplanar plate capacitors
on a film substrate and a capacitive touch integrated circuit (IC). The performance of the developed
sensors is also evaluated.

2. Materials and Methods

2.1. Sensor Development

One of the key technologies for a low-cost soil moisture sensor is film printing. The recent
development of printed circuits on film is drawing attention because it has a lower cost than regular
electrical circuits and, in particular, it allows for inexpensive, easy, and rapid prototyping such that
the cost and time of development can be reduced [32]. A soil moisture sensor can be also made
on a film [33,34]. A variety of sensor designs can be examined easily and economically with this
technology. Thus, the best design for a soil moisture sensor adaptable to various environments can be
developed with low cost. In this study, we developed the copper film substrate shown in Figure 1. The
circuit was prepared by etching copper on a polyethylene terephthalate (PET) film. We implemented
soil and soil-surface temperature measurement functions on the sensor. Thus, the circuit included
sensing parts for soil moisture, sensing parts for temperature, and wiring parts. The circuit had
three soil moisture sensing parts to measure the electrical capacitance of soil at 10, 20, and 30 cm
Sensors 2016, 16, 1292 3 of 13

Sensors 2016, 16, 1292 3 of 13

14
in deeper soil layers can be measured by developing longer film circuit. Each sensing part consisted
of
in two wide
deeper soilbars,
layerswith
canwidth and length
be measured of 25 and longer
by developing 55 mm,film respectively.
circuit. Each There
sensing waspart
a 1 consisted
mm gap
between
of two wide bars, with width and length of 25 and 55 mm, respectively. There was a 1 mm was
depths. the
These two bars.
depths wereThese two
selected bars
to work
evaluate as
soila capacitor.
moisture in Therefore,
the root zonethe forsensing
many area
crops. Soil
gap
25 mm
moisture × 55
in mm
deeper× 2 (two
soil bars)
layers =
can2750
be mm 2 (=27.5
measured bycm 2). The sensing
developing
between the two bars. These two bars work as a capacitor. Therefore, the sensing area was longer parts
film for temperature
circuit. Each were
sensing also
part
located
25 mm at
consisted 10,
× 55of mm20, and
two wide
× 30bars,
2 (two cm depth,
with
bars) andmm
= width
2750 there
and was an
length
2 (=27.5 ofextra
cm 225 and
). The temperature
55 mm,parts
sensing sensing
respectively. part 2 cmwas
There
for temperature above the
a 1 also
were mm
soil
gap surface.
between Thethewiring
two part
bars. extended
These two to the
bars end
work of the
as a film and
capacitor. was
located at 10, 20, and 30 cm depth, and there was an extra temperature sensing part 2 cm above the connected
Therefore, the to a measurement
sensing area was
circuit. × 55film
The × 2rolled 2 2and
25 mm
soil surface. Thewas
mm wiring(two on extended
part a PVC
bars) = 2750pipeto whose
mm endouter
the(=27.5 cmthe
of inner
). film
The anddiameters
sensingwasparts were
for
connected 32toand
temperature 25 mm,
wereand
a measurement also
was covered
located
circuit. at 10,
The by20,awas
film 50 μm
and 30 cm
rolled (after
on ashrinking)
depth,PVCandpipe PET
there
whose heat
was anshrinking
outer extra film.
andtemperature
inner Before
diameters covering
sensing
werepart it2 with
32 and cm the PET
25above
mm, the
and
film,
soil NTC
surface. thermistor
The wiringtemperature
part sensors
extended to (NCP18WF104J03RB,
the end of the film and Murata
was covered by a 50 μm (after shrinking) PET heat shrinking film. Before covering it with the PETwas manufacturing
connected to a Company,
measurement
Ltd.,
film, Kyoto,
circuit.
NTC Japan)
Thethermistor
film was were
rolledpasted
on a on
temperature PVCeach
pipe
sensorstemperature-sensing
whose outer and inner
(NCP18WF104J03RB, part.diameters
An edge
Murata of the32
were
manufacturingPVC
andcolumnmm,was
25Company,and
filled
was with
covered a silicone
by a 50 sealant
µm (afterfor waterproofing.
shrinking) PET A
heat photograph
shrinking of
film.
Ltd., Kyoto, Japan) were pasted on each temperature-sensing part. An edge of the PVC column wasthe developed
Before covering sensors
it with is
theshown
PET in
film,
Figure
NTC 2.
thermistor temperature sensors (NCP18WF104J03RB, Murata
filled with a silicone sealant for waterproofing. A photograph of the developed sensors is shown in manufacturing Company, Ltd.,
Kyoto, 2.
Figure Japan) were pasted on each temperature-sensing part. An edge of the PVC column was filled
with a silicone sealant for waterproofing. A photograph of the developed sensors is shown in Figure 2.

Figure 1. Schematic of copper circuit printed on the film substrate.

Figure 1. Schematic of copper circuit printed on the film substrate.
Figure 1. Schematic of copper circuit printed on the film substrate.

Developedsoil
Figure2.2.Developed
Figure soilmoisture
moisturesensors.
sensors.

Figure 2. Developed soil moisture sensors.

The other
The other key
key technology
technology isis aa capacitive
capacitive touchtouch IC
IC for
for touch
touch sensors.
sensors. To To date,
date, capacitive
capacitive soilsoil
moisture
moisture
The measurement has
has been
other key technology beenismade
made with
withrelatively
a capacitive touch high
relativelyIChigh
forfrequency,
frequency,
touch i.e.,i.e.,
sensors. 70 To
MHz or higher,
70 date,
MHz in order
or higher,
capacitive in
soil
to obtain
order good
to obtain
moisture accuracy [27],
good accuracy
measurement has beenand the
[27],
made soil
andwith moisture sensors
the relatively
soil moisture are
highsensors expensive
frequency, because
are expensive
i.e., 70 MHz of the
because relatively
of the
or higher, in
high-frequency
relatively capacitance
high-frequency measurement
capacitance circuit
measurement that is
circuit required.
that is The
required.
order to obtain good accuracy [27], and the soil moisture sensors are expensive because of the recent
The development
recent developmentof the
capacitive
of
relatively touch IC
the capacitive
high-frequencyenables
touch IC us to measure
enables
capacitance us to electrical capacitance
measurecircuit
measurement electrical without investing
that iscapacitance
required. Thewithout
recentindevelopment
complicated
investing in
circuits.
complicatedAs the capacitive
circuits.
of the capacitive As the
touch touch IC
ICcapacitive is
enables us being
touch embedded
IC is being
to measure in a variety
embedded
electrical of devices,
in a variety
capacitance such as smartphones
of devices,
without such as
investing in
and tablets,
smartphones many low-cost capacitive touch ICs that can be operated
complicated circuits. As the capacitive touch IC is being embedded in a variety of devices, with
and tablets, many low-cost capacitive touch ICs that with
can a
be microcomputer
operated sucharea
as
available.
microcomputerAlthough
are the measurement
available. Although frequency
the possible
measurement with those
frequency
smartphones and tablets, many low-cost capacitive touch ICs that can be operated with a ICs is
possible lower than
with with
those the
ICs iscurrent
lower
commercial
than with thesoil
microcomputer moisture
current
are sensors
commercial
available. (e.g.,
soil
Although 50 kHz–3
moisture
the MHz),
sensors
measurement the
(e.g., cost
50
frequency reduction
kHz–3 MHz),
possible bythe
withusing
cost the
those capacitive
reduction
ICs by
is lower
touch
using IC
the is significant.
capacitive Thus,
touch IC using
is a capacitive
significant. Thus,touch
usingIC is
a a good
capacitive choice
touch
than with the current commercial soil moisture sensors (e.g., 50 kHz–3 MHz), the cost reduction by forICdeveloping
is a good a low-cost
choice for
using the capacitive touch IC is significant. Thus, using a capacitive touch IC is a good choice for
Sensors 2016, 16, 1292 4 of 13
Sensors 2016, 16, 1292 4 of 14
developing a low-cost soil moisture sensor. In this study, a proximity capacitive touch sensor
controller MPR121 (NXP Semiconductors N.V., Eindhoven, The Netherlands) was used for
soil moisture
measuring thesensor. In this
capacitance, andstudy, a proximity capacitive
a microcomputer touchSemiconductors
LPC1114 (NXP sensor controller MPR121
N.V., (NXP
Eindhoven,
Semiconductors N.V., Eindhoven, The Netherlands) was used for measuring the capacitance,
The Netherlands) was used for controlling the MPR121 and for data logging. The MPR121 measures and
a microcomputer
the LPC1114
electric capacitance C [F](NXP
by theSemiconductors
constant currentN.V., Eindhoven,
source TheThe
method [35]. Netherlands)
method runswas used for
a constant
controlling the MPR121 and for data logging. The MPR121 measures the electric capacitance
direct current I [A] into the capacitor during a unit time period t [s]. There is a relationship between C [F]
by the constant current source
the C, I, t, and voltage V [V]: method [35]. The method runs a constant direct current I [A] into the
capacitor during a unit time period t [s]. There is a relationship between the C, I, t, and voltage V [V]:
V = I·t/C (1)
With Equation (1), V was processed by V =the I·t/C
analog–digital conversion to determine C. (1) A
measurement frequency of approximately 62 kHz was used with the developed sensor. This is the
With Equation (1), V was processed by the analog–digital conversion to determine C.
default frequency in the MPR121 (MPR121 allows us to choose among 31, 62, and 125 kHz). The
A measurement frequency of approximately 62 kHz was used with the developed sensor. This is
reason we decided to use the 62 kHz frequency was because it offers the finest resolution of voltage
the default frequency in the MPR121 (MPR121 allows us to choose among 31, 62, and 125 kHz).
measurement. The capacitances measured in soils with the developed sensor ranged between
The reason we decided to use the 62 kHz frequency was because it offers the finest resolution of voltage
40 and 1000 pF in a preliminary experiment. Given this capacitance range, the voltage measurement
measurement. The capacitances measured in soils with the developed sensor ranged between 40 and
resolution was higher at 62 kHz than at the other frequencies. It has been reported that higher
1000 pF in a preliminary experiment. Given this capacitance range, the voltage measurement resolution
frequency is preferred in order to reduce the influence of soil temperature and electrical conductivity
was higher at 62 kHz than at the other frequencies. It has been reported that higher frequency is
[27]; however, those considerations were made for frequencies in the MHz range. The difference
preferred in order to reduce the influence of soil temperature and electrical conductivity [27]; however,
between 62 kHz and 125 kHz is small and we concluded that it is more important to obtain a finer
those considerations were made for frequencies in the MHz range. The difference between 62 kHz and
resolution than to use a slightly higher frequency.
125 kHz is small and we concluded that it is more important to obtain a finer resolution than to use a
The developed soil moisture sensors can work as a sensor network system by providing them
slightly higher frequency.
with a wireless communication function. The sensor network system has multiple sensors with an
The developed soil moisture sensors can work as a sensor network system by providing them
embedded wireless communication module that are distributed in a field, and their measured data
with a wireless communication function. The sensor network system has multiple sensors with an
are efficiently collected through wireless communication [36,37]. Figure 3 shows a scheme of this
embedded wireless communication module that are distributed in a field, and their measured data are
sensor network system. The sensor network system consisted of slave nodes, which were the
efficiently collected through wireless communication [36,37]. Figure 3 shows a scheme of this sensor
developed soil moisture sensors that measure the electric capacitance of the soil at multiple points,
network system. The sensor network system consisted of slave nodes, which were the developed soil
and a master node, which transfers the measured data into a web server. The details of each node are
moisture sensors that measure the electric capacitance of the soil at multiple points, and a master node,
shown in Figure 4.
which transfers the measured data into a web server. The details of each node are shown in Figure 4.

Master node
Wireless
Farm field communication

・ ・・ ・ ・

Slave nodes

・ ・・ ・ ・

Figure 3. Scheme
Scheme of the sensor network system with the developed soil moisture sensors.

The slave node consisted of a microcomputer, a battery box, and a communication unit. The
microcomputer and the soil moisture-sensing part were connected by a coaxial cable. The power
supply (6 V DC) consisted of four D cell batteries connected in series. The electrical requirement of
adapter for the power supply. The Android smartphone had an inexpensive SIM card embedded,
which collects data from the slave nodes and sends the data to the internet server via cellphone
communication. A used Nexus 5 (LG Electronics, Seoul, Korea) was purchased and used for the
sensor network system in this study. Other low-cost Android smartphones can be an alternative. The
measured capacitance data was also stored on the SD card installed in the microcomputer. The
Sensors 2016, 16, 1292 5 of 14
master node collected data from the slave nodes and sent them to the server every hour.

Communication (a) Master node

unit
Send the measured
data to the server
Box
Communication Android
unit Smart phone
(Tocos) AC
Box
adopter
Converter
board AC100 V
Cable

To AC power

Communication (b) Slave node

unit

Battery box
Microcomputer
Communication Unit
Box Antenna

Box
Communication
unit
Cable
(to sensor)
To soil moisture sensor

Figure 4.
Figure Details of
4. Details of (a)
(a) master
master node
node and
and (b)
(b) slave
slave node
node of
of the
the sensor
sensor network
network system.
system.

2.2. Sensor Calibration

The slave node consisted of a microcomputer, a battery box, and a communication unit.
The The
microcomputer
measured and the soilcapacitance
electrical moisture-sensing
of thepart
soilwere connected
at each depthbyneeds
a coaxial
to cable. The power
be converted to
supply (6 V DC) consisted of four D cell batteries connected in series. The
volumetric water content (VWC). VWC is defined as water volume per unit soil volume, and thus, electrical requirement
of the
the unitslave nodeisismapproximately
of VWC 1 mW, and obtained
3·m−3. Usually, empirically it is expected that such
equations the batteries
as Topp’s will last for[11]
equation 1 year.
are
The
used.master
In thisnode
study,consisted of an Android
we developed our own smartphone,
empiricalaequations
communication unit, and aexperiment.
by a calibration converter board.
The
The communication
developed unit in
soil moisture both master
sensor and slave
was placed nodes
in a 50 cm used
long aacrylic
Tocos column
strong module
whose (Tokyo Cosmos
inner diameter
Electric
was Co., The
10 cm. Ltd.,bottom
Tokyo, Japan) in the 2.4was
of the column GHzcovered
band, which
with aallows
fabricdata transmission
membrane. The over a distancethe
gap between of
1sensor
km. The Android smartphone requires 5 V DC. In this study, we used 100 V AC
and the column was filled with soil sampled at a greenhouse field in Ibaraki prefecture,and an AC adapter for
the power
Japan. Thesupply.
soil wasThe Android smartphone had
organic-matter-enriched an inexpensive
volcanic SIM card
ash soil (sandy loam),embedded, which collects
and was packed with a
data from the slave nodes and sends the data to the internet server via cellphone
bulk density of 0.8 Mg·m . Three commercialized soil moisture sensors, 5TE (Decagon Devices, Inc.,
−3 communication.
A used Nexus
Pullman, WA,5USA),
(LG Electronics,
were insertedSeoul, Korea)the
through wasside
purchased
wall of and used forat
the column thedepths
sensorofnetwork system
10, 20, and 30
in this study. Other low-cost Android smartphones can be an alternative. The
cm from the soil surface. The column was placed in a container, which had approximately 5 cm measured capacitance
data waswater,
ponded also stored on the was
and water SD card installed
infiltrated in the
into themicrocomputer. The master
soil from the bottom node collected
by capillary force. data
The
from the slave nodes and sent them to the server every hour.
ponded water height was manually maintained at 5 cm during the entire experiment.
2.2. Sensor Calibration
The measured electrical capacitance of the soil at each depth needs to be converted to volumetric
water content (VWC). VWC is defined as water volume per unit soil volume, and thus, the unit
of VWC is m3 ·m−3 . Usually, empirically obtained equations such as Topp’s equation [11] are used.
In this study, we developed our own empirical equations by a calibration experiment. The developed
soil moisture sensor was placed in a 50 cm long acrylic column whose inner diameter was 10 cm.
The bottom of the column was covered with a fabric membrane. The gap between the sensor and
the column was filled with soil sampled at a greenhouse field in Ibaraki prefecture, Japan. The soil
was organic-matter-enriched volcanic ash soil (sandy loam), and was packed with a bulk density of
0.8 Mg·m−3 . Three commercialized soil moisture sensors, 5TE (Decagon Devices, Inc., Pullman, WA,
Sensors 2016, 16, 1292 6 of 14

Sensors 2016, 16, 1292 6 of 13

USA), were inserted through the side wall of the column at depths of 10, 20, and 30 cm from the soil
surface.
TheThe column
electrical was placedand
capacitance in athe
container,
VWC at which had approximately
the three 5 cm ponded
depths were measured withwater,
both and
the
water was infiltrated
developed sensor and into
thethe5TEs.
soil from
The the
5TEbottom
has beenby capillary
used in force.
variousThe ponded
studies aswater height was
an accurate soil
manually
moisture maintained at 5 cm for
sensor developed during the entire
research use. experiment.
Therefore, the VWC measured with the 5TE was
The electrical capacitance and the VWC at
considered as a reference at each depth. Water infiltration the threecontinued
depths wereuntilmeasured with both
the capacitance at thethe
10
developed
cm depth sensor
became andconstant.
the 5TEs.ThisThe 5TE has beenwas
procedure usedrepeated
in variousfive
studies
timesas an accurate
with soil moisture
the same sensor.
sensor developed
Preliminary for research
capacitance use. Therefore,
measurements using the
the VWC
sensormeasured
in air andwith theindicated
water 5TE was considered as a
that the sensor
reference at each depth.
output variability Waterwhich
was small, infiltration continued
justified using theuntil
samethesensor
capacitance at the 10the
for repeating cmprocedure.
depth becameThe
constant.
approximate Thisequations
procedure forwas
therepeated five times
three depths were with the same
obtained sensor.
by the Preliminary
least square capacitance
method from the
measurements using thethe
relationships between sensor in air and and
capacitance waterVWC.
indicated
Thisthat the sensorprocedure
calibration output variability
is easy, was
fast,small,
and
which justified
sufficiently using the same sensor for repeating the procedure. The approximate equations for the
reliable.
three depths were obtained by the least square method from the relationships between the capacitance
2.3. Sensor
and VWC. Evaluation
This calibration procedure is easy, fast, and sufficiently reliable.
The performance
2.3. Sensor Evaluation of the developed sensor was tested at various fields. In this study, we
introduced the data from a grape (Vitis vinifera L.) field at the Institute of Sustainable
The performance
Agro-ecosystem Services,of the
the University
developed of sensor
Tokyowas tested Tokyo,
(Tanashi, at various fields.
Japan) In this
and from study, we
a greenhouse
introduced
field wherethe data from
mizuna a grape
(Brassica var. vinifera
rapa(Vitis L.) field
laciniifolia) was at the Institute
grown of Sustainable
in the Ibaraki Agro-ecosystem
Prefecture. The mizuna
Services, the University of Tokyo (Tanashi, Tokyo, Japan) and from a greenhouse
was planted in the greenhouse during the experiment. The soils at both fields were volcanic ash field where mizuna
soil.
(Brassica rapa var. laciniifolia) was grown in the Ibaraki Prefecture. The mizuna
Four sensors (A, B, C, and D) were installed at the grape field. As the grape field was furrowed, was planted in two
the
greenhouse
of them were during
insertedtheinto
experiment.
a ridge (A, The
B) soils
and theat both
othersfields
werewere volcanic
inserted intoash soil. Four
a furrow sensors
(C, D). Nine
(A,
sensors were installed at the mizuna field to capture the spatial variability of soil moisture inwere
B, C, and D) were installed at the grape field. As the grape field was furrowed, two of them the
inserted
greenhouseinto(a–i).
a ridge
The (A,
sizeB)of and the others were
the greenhouse and the inserted intoofathe
locations furrow (C, D). Nine
soil moisture sensorssensors were
are shown
installed
in Figureat5.the mizuna
Sensor field to capture
installations were the spatial
carried variability
out by digging of soil moisture in hole
a 35-cm-deep the greenhouse
with an auger,(a–i).
The size of the greenhouse and the locations of the soil moisture sensors are shown
placing the sensor in the hole, and filling the space with soil from the field. Because the soil was in Figure 5. Sensor
installations
aggregated, itwerewascarried
crushed out by digging
manually a 35-cm-deep
before hole with
filling the space. Thean auger, placing
experimental the sensor
period was fromin the
15
hole, and filling the space with soil from the field. Because the soil was aggregated,
October to 19 December 2015 for the grape field, and from 12 September to 17 October 2015 for the it was crushed
manually before filling the space. The experimental period was from 15 October to 19 December 2015
mizuna field.
for the grape field, and from 12 September to 17 October 2015 for the mizuna field.

50 m

a d g
Entrance
N

1.5 m
5m

b e h
c f 15 m i 10 m

Greenhouse
M aster node Slave node

Figure 5. Dimensions of the greenhouse and locations of the master and slave nodes.
Figure 5. Dimensions of the greenhouse and locations of the master and slave nodes.
3. Results and Discussion
3. Results and Discussion
3.1. Sensor Calibration
3.1. Sensor Calibration
Figure 6 shows the relationship between the VWC measured with the 5TE and the electrical
Figure measured
capacitance 6 shows the withrelationship
the developed between the VWC
soil moisture measured
sensors. Fivewith
curvesthewith
5TEdifferent
and the colors
electrical
are
capacitance measured with the developed soil moisture sensors. Five curves with
shown at each depth because the experiment was repeated five times with a same sensor. The VWC different colors are
shown at each
increased from depth
0.03 tobecause
0.35 m the
3 −
·m experiment
3 was repeated
, and the capacitance five between
varied times with100a and
same sensor.
1000 The VWC
pF depending
increased from 0.03 to 0.35 m 3·m−3, and the capacitance varied between 100 and 1000 pF depending
on soil moisture. There was a strong positive correlation between VWC and capacitance. Thus, the
on soil moisture.
developed sensorThere wastoa detect
was able strongchanges
positive incorrelation
VWC. Thebetween VWC
five curves didandnotcapacitance.
match well Thus, the
with one
developed
another, sensor
except was30
at the able
cmto detectThe
depth. changes in VWC.
main reason forThe
thefive curvesindid
variation not match
curves well
at the 10 cmwith
and one
the
another,
20 except
cm depths at the
may 30 cm
be that depth.
water The main
infiltrated fromreason for the reached
the bottom variationthe
in 5TE
curvesandatthe
thedeveloped
10 cm and soil
the
20 cm depths may be that water infiltrated from the bottom reached the 5TE and the developed soil
moisture sensor at different times. The 5TE and the sensing part of the developed soil moisture
sensor have different sizes and different sampling volumes. The 5TE has a 5-cm-long and
Sensors 2016, 16, 1292 7 of 14

Sensors 2016, 16, 1292 7 of 13

moisture sensor at different times. The 5TE and the sensing part of the developed soil moisture sensor
2.5-cm-wide sensing part, which is smaller than that of the developed soil moisture sensor (the
have different sizes and different sampling volumes. The 5TE has a 5-cm-long and 2.5-cm-wide sensing
developed sensor has a 5.5-cm-long and 5-cm-wide sensing part). It is also assumed that these
part, which is smaller than that of the developed soil moisture sensor (the developed sensor has a
sensors have different sampling volumes since the sampling volume depends on sensor design and
5.5-cm-long and 5-cm-wide sensing part). It is also assumed that these sensors have different sampling
soil conditions [38]. In addition, it was visually observed that the wetting front of the soil was not
volumes since the sampling volume depends on sensor design and soil conditions [38]. In addition,
horizontal during the experiment. Water infiltrated faster where the soil is slightly loose due to a
it was visually observed that the wetting front of the soil was not horizontal during the experiment.
larger hydraulic conductivity than where soil is dense. Thus, the water reached the sampling volume
Water infiltrated faster where the soil is slightly loose due to a larger hydraulic conductivity than
of one of the sensors faster than that of the other. As the elapsed time until water reaches the sensing
where soil is dense. Thus, the water reached the sampling volume of one of the sensors faster than that
parts of the sensors at the 30 cm depth was much shorter than at the other depths, the five curves at
of the other. As the elapsed time until water reaches the sensing parts of the sensors at the 30 cm depth
the 30 cm probably became similar.
was much shorter than at the other depths, the five curves at the 30 cm probably became similar.

0.35
0.30 10 cm 20 cm 30 cm
0.25
VWC (m3∙m-3)

0.20
0.15 1 1
1
0.10 2 2 2
3 3 3
0.05 R 2 = 0.918 4 R 2 = 0.945 4 R 2 = 0.926 4
5 5 5
0.00
0 250 500 750 1000 0 250 500 750 1000 0 250 500 750 1000
Capacitance (pF) Capacitance (pF) Capacitance (pF)

Figure 6. Relationship
Figure 6. Relationship between
between volumetric
volumetric water
water content
content (VWC)
(VWC) and
and electrical
electrical capacitance
capacitance obtained
obtained
from the calibration experiments. The different colors and numbers indicate the repeated
from the calibration experiments. The different colors and numbers indicate the repeated count. count.
The
The black lines are the approximate equations.
black lines are the approximate equations.

to determine
As it was difficult to determine which curve is thethe most
most appropriate,
appropriate, we we derived
derived approximate
approximate
equations from all sampled
sampled data.
data. Based
Basedononthe
thecurve
curveshape,
shape,we weconcluded
concludedthatthata alinear
linearfunction, a
function,
aquadratic
quadraticfunction, andand
function, a cubic function
a cubic are appropriate
function for the
are appropriate for10the
cm,1020cm,
cm,20and cm,30 and
cm depths,
30 cm
respectively:
depths, respectively:
θ = 2.8 × 10-4−4C + 3.2 × 10−2-2 at 10 cm depth (2)
θ = 2.8 × 10 C + 3.2 × 10 at 10 cm depth (2)
θ = −3.7 × 10-7 C2 + 7.1 × 10-4 C − 3.5 × 10-2 at 20 cm depth (3)
θ = −3.7 × 10−7 C2 + 7.1 × 10−4 C − 3.5 × 10−2 at 20 cm depth (3)
θ = 8.8 × 10-10 C3 − 2.0 × 10-6 C2 + 1.6 × 10-3 C − 1.4 × 10-1 at 30 cm depth (4)
θ = 8.8θ×indicates
In these equations, 10−10 C3 −the2.0volumetric
× 10−6 C2 + 1.6 × 10
water
−3 C − 1.4 × 10−1 at 30 cm depth
content and the unit of C is pF. The coefficients (4)
of determination, R2 , for θ
In these equations, 10indicates
cm, 20 cm, theand 30 cm depth
volumetric water were 0.918,and
content 0.945,
theand 0.926,
unit of Crespectively.
is pF. The
The sensor accuracies
coefficients corresponding
of determination, R2, forto10thecm,95%20confidence
cm, and 30 intervals
cm depthof the model
were equations
0.918, were0.926,
0.945, and 0.05,
0.03, and 0.02 m 3 ·m−3 , respectively. The reason that each depth had a different curve shape may be
respectively. The sensor accuracies corresponding to the 95% confidence intervals of the model
the effect ofwere
equations the wiring partand
0.05, 0.03, on the
0.02film substrate.
m3·m The wiring
−3, respectively. Thepart in Figure
reason 1 alsodepth
that each works as a capacitor
had different
that
curve measures
shape maya minor
be thecapacitance depending
effect of the on the
wiring part onsoil
themoisture on the wires.
film substrate. The length
The wiring of Figure
part in the wire1
is proportional to the depth. Therefore, the relationship between VWC and
also works as a capacitor that measures a minor capacitance depending on the soil moisture on C was linear at the 10 the
cm
depth
wires. because
The lengththere waswire
of the not is
a significant
proportional capacitance on the
to the depth. short wire.
Therefore, However, the
the relationship relationship
between VWC
between
and C was VWC andatCthe
linear became
10 cmmoredepth complex
becausewith thewas
there larger
notinfluence of the
a significant capacitance
capacitance onmeasured
the short
on theHowever,
wire. wiring part theat relationship
the 20 and 30betweencm depths. VWCThisandmight be problematic
C became in fieldwith
more complex measurements,
the larger
and in future
influence studies
of the it is necessary
capacitance measuredtoon find
theawiring
method to at
part eliminate the 30
the 20 and effect of the wiring
cm depths. part on
This might be
the capacitance measurement. Hereafter, the capacitance measured with the developed
problematic in field measurements, and in future studies it is necessary to find a method to eliminate sensors was
converted to the
the effect of VWC with part
wiring the equations.
on the capacitance measurement. Hereafter, the capacitance measured
with the developed sensors was converted to VWC with the equations.

3.2. Sensor Performance

Figure 7 shows the temperature, the VWC measured with the developed sensors at the grape
field in Tokyo, and the precipitation rate. The measured VWC initially maintained relatively small
Sensors 2016, 16, 1292 8 of 14

3.2. Sensor Performance

Figure 7 shows the temperature, the VWC measured with the developed sensors at the grape field
in Tokyo, and the precipitation rate. The measured VWC initially maintained relatively small values
(<0.15 m3 ·m−3 ) after the sensor installation on 14 October, and then increased by 0.01–0.08 m3 ·m−3 due
to precipitation on 16 and 17 October. The VWC kept increasing after 17 October until 2 November even
though there was not a large amount of precipitation. Relative to the value measured on 17 October,
the VWC measured just before the precipitation on 2 November had increased by approximately
0–0.03 m3 ·m−3 . This is due to the improvement in the contact between sensors and soil, i.e., the soil
disturbed by the sensor installations settled down slowly, and thus the soil around the capacitors
became dense. The contact improvements ceased by the rainfall on 2 November, after which the
measured VWC ranged between 0.11 m3 ·m−3 and 0.26 m3 ·m−3 . This indicates that there is some
period, two weeks in this case, required for the sensors to start measuring VWC reliably. In addition,
the VWC measured with the sensor D suddenly decreased by 0.01–0.03 m3 ·m−3 at all depths on
31 October 2015. Weeding was performed at the grape field at this moment, and it was found that a
worker had touched the sensor head exposed on the soil surface. It is thought that a space between the
sensor and the soil was created, and the contact between them deteriorated by the incident. Clearly,
the contact between the sensor surface and the soil is crucial for the developed soil moisture sensors.
The measured VWC showed increases due to precipitation and decreases due to evapotranspiration
during the daytime after 2 November. The fluctuation of VWC was between 0.11 m3 ·m−3 and
0.26 m3 ·m−3 . There were missed data from 14 November to 17 November due to urgent maintenance
of the data logging server. At that time, the system did not have a data resend function in case the data
transmission failed (the function was included after this incident). The measured data were stored in
a SD card in the slave node, so the missing data can be complemented by collecting data at the field
from the SD card. Except for the period with missing data, the VWC measured with the developed
sensor was reasonable and similar to those found in other studies [20,30]. The other remark is that
the VWC sometimes dropped suddenly to values less than 0.10 m3 ·m−3 . This phenomenon was most
frequently found with sensor D after 3 December. This may be associated with a problem of circuit
contact around the capacitance measurement IC. Possibly electrical current leaked somewhere in the
circuit and the recorded capacitances were smaller than the real values. The collection rate of regular
data during the experimental period was 97.0%.
The measured VWCs tended to be larger with depth, i.e., θ at 10 cm < θ at 20 cm < θ at 30 cm.
This is reasonable because the shallow soil layers easily lose water because of their interaction with the
atmosphere and are more likely to be loose, so that they cannot hold as much water as the deeper layers.
It was expected that the VWCs measured in ridges would be smaller than those measured in furrows
because more soil surface is exposed to the atmosphere and the soil is looser in a ridge. However,
no clear variation due to differences in location was observed. There might be two reasons for this:
(1) the individual differences of sensor output masked the variations in VWC at different locations; and
(2) the difference in soil compaction in a ridge location was diminished by the disturbance of soil when
the sensors were installed. As the variability of sensor output in air and water was not significant in
the preliminary experiment, the individual differences of sensor outputs must be associated with the
variability of the contact between each sensor’s capacitors and the surrounding soil. Hence, having
good and similar contact conditions at each point is essential to determining the effect of location
differences on soil moisture.
Sensors 2016, 16, 1292 9 of 13

measured VWC values at most points and depths increased by a maximum of 0.05 m3·m−3 after the
irrigation events. Besides the increases due to irrigation, decreases of VWC due to
Sensors 2016, 16, 1292
evapotranspiration 9 of 14
were also observed. Therefore, the developed soil moisture sensor satisfactorily
detected dynamic changes in VWC.
30
Above surface 10 cm
25

Temperature (oC)
20 cm 30 cm
20
15
10
5 (i) Tem perature
0
0.3
(ii) VW C at A & B (ridg e)
VWC (m3∙m−3)

0.2

0.1
A 10cm A 20cm A 30cm B 10cm B 20cm B 30cm

0.30
(iii) VW C at C & D (furrow )
VWC (m3∙m−3)

0.2

0.1

C 10cm C 20cm C 30cm D 10cm D 20cm D 30cm

0
80
Precipitation (mm∙d−1)

(iv) Precipitation
60

40

20

0
10/15 10/22 10/29 11/5 11/12 11/19 11/26 12/3 12/10 12/17

Figure 7.7.Temperatures
Temperatures above thethe
above surface and at
surface and10,at
20,10,
and20,
30 and
cm depths
30 cmmeasured with the developed
depths measured with the
sensor A (i),
developed volumetric
sensor water content
A (i), volumetric water(VWC)
contentmeasured in ridges
(VWC) measured in (ii),
ridgesand
(ii),inand
furrows (iii), (iii),
in furrows and
precipitation (iv) at
and precipitation theatgrape
(iv) field field
the grape from from
25 October to 19 to
25 October December 2015.2015.
19 December

Figure 8 shows VWCs measured with the developed sensors at the mizuna field in Ibaraki. The left
(i) Sensor a, c, e, and f (ii) Sensor g, h, and i
panel shows the VWC with sensors a, c, e, and f, located at the middle or north of the greenhouse,
0.30
while the right panel shows the VWC with sensors g, h, and i, located at the south. The sensors b and
VWC (m3∙m-3)

d began malfunctioning immediately after irrigation by the sprinklers installed at the ceiling of the
greenhouse
0.20 on 15 September (data not shown). It was found that the communication unit circuits of
the sensors b and d were broken owing to water seepage. After sensor installation on 11 September, the
measured VWC increased continuously until 20 September. Through this time interval, the measured
VWC0.10
increased by 0.02–0.07 m3 ·m−3 compared to the initial value. It should be noted that there was
an irrigation event
a 10cm on 15 September.
a 20cm a 30cm c 10cm These
c 20cm observations
c 30cm indicate
g 10cm gthat
20cm the contact
g 30cm h 10cmbetween
h 20cm sensor
e 10cm e 20cm e 30cm f 10cm f 20cm f 30cm h 30cm i 10cm i 20cm i 30cm
capacitors and soil improved for 10 days, as the sensors at the grape field did. Irrigation was performed
0.00
twice during the experimental period, on 15 and 299/12 September. The 9/26 measured VWC10/10values at most
9/12 9/19 9/26 10/3 10/10 10/173 −39/19 10/3 10/17
points and depths increased by a maximum of 0.05 m ·m after the irrigation events. Besides the
increases due
Figure to irrigation,
8. Volumetric decreases
water contentof VWCmeasured
(VWC) due to evapotranspiration
with the developedwere also observed.
soil moisture sensors Therefore,
at the
the developed soil moisture sensor satisfactorily detected dynamic changes
mizuna green house field from 12 September to 17 October 2015. The left panel (i) shows VWC at in VWC.
locations a, c, e, and f, and the right panel (ii) shows VWC at locations g, h, and i.
 0
10/15 10/22 10/29 11/5 11/12 11/19 11/26 12/3 12/10 12/17

Figure 7. Temperatures above the surface and at 10, 20, and 30 cm depths measured with the
developed sensor A (i), volumetric water content (VWC) measured in ridges (ii), and in furrows (iii),
Sensorsand
2016, 16, 1292
precipitation (iv) at the grape field from 25 October to 19 December 2015. 10 of 14

(i) Sensor a, c, e, and f (ii) Sensor g, h, and i

0.30
VWC (m3∙m-3)

0.20

0.10
a 10cm a 20cm a 30cm c 10cm c 20cm c 30cm g 10cm g 20cm g 30cm h 10cm h 20cm
e 10cm e 20cm e 30cm f 10cm f 20cm f 30cm h 30cm i 10cm i 20cm i 30cm

0.00
9/12 9/19 9/26 10/3 10/10 9/12
10/17 9/19 9/26 10/3 10/10 10/17

Figure
Figure 8. Volumetric water
8. Volumetric water content
content (VWC)
(VWC) measured
measured with
with the
the developed
developed soil
soil moisture
moisture sensors
sensors at
at the
the
mizuna green house
mizuna green house field
field from
from 12
12 September
September toto 17
17 October
October 2015. The left
2015. The left panel
panel (i)
(i) shows
shows VWC
VWC at at
locations
Sensorslocations a, c, e, and f, and the right panel (ii) shows VWC at locations g, h, and
a, c, e, and f, and the right panel (ii) shows VWC at locations g, h, and i.
2016, 16, 1292 i. 10 of 13

VWCs
VWCs in in the right panel
panel of Figure 8 (g, h, i) were smaller and changed more dynamically than
those
those in the left panel.
panel. For example, the soil soil moistures
moistures at at 10
10 cm
cm at atggandandhhwere were0.05–0.15
0.05–0.15mm ·m−−33
3 ·3m

lower
lower thanthan those
those at other sensor
sensor locations,
locations, i.e.,
i.e., a,
a, c, e, f, and i. The daily decreases
decreases of VWC after the
irrigation on 3 ·m−3− 3 , while
on 29 29 September
September at at gg and
and hhwere
weresometimes
sometimesclose closeto to0.02
0.02m m3·m , while those
thoseat at
locations
locations a,
c, e, f, and i were always less than 0.01 m 3·m 3 − 3
−3. Figure 9 shows the increase
a, c, e, f, and i were always less than 0.01 m ·m . Figure 9 shows the increase rates of VWC after rates of VWC after the
irrigation on 29
the irrigation on September.
29 September.At At10 10
cmcmdepth,
depth, thetheVWCVWC atat
g gandandhhincreased
increaseddynamically,
dynamically, 5.3% 5.3% and
4.7%, respectively, but not at the other locations. Only the VWC at h showed
respectively, but not at the other locations. Only the VWC at h showed a large increase (4.2%) a large increase (4.2%) at
at
2020cmcm depth,
depth, butbut there
there was
was nono dynamic
dynamic change
change in in
VWCVWC at at
thethe
3030 cmcm depth.
depth. AsAsthethe locations
locations ggandandh
h caughtplenty
caught plentyofofsunshine
sunshinecompared
comparedtotothe theother
otherplaces,
places,thetheevapotranspiration
evapotranspiration at these points was
enhanced
enhanced and and thethe soil was relatively dry. This condition condition probably
probably caused
caused the dynamicdynamic increase
increase in
VWC at g and h during the irrigation. Therefore, Therefore, there there were relatively
relatively large differences (more than
0.05 m33·m
0.05 ·m−3−)3in
) insoil
soilmoisture
moisturewithin
withinthe
themizuna
mizunagreenhouse
greenhousefield field and
and thethe developed
developed soil soil moisture
sensor was satisfactory able to detect such variations at the different locations.

10 cm dep th

a 1.1% d - g 5.3%
N

b - e 0.9% h 4.7%
c 0.9% f 0.6% i 1.4%
20 cm dep th
a 0.5% d - g 0.7%
N

b - e 0.6% h 4.2%

c 0.0% f 0.5% i 0.1%

30 cm dep th

a 0.6% d - g 0.1%
N

b - e 0.9% h 1.4%

c 0.0% f 0.8% i 0.0%

Figure
Figure 9.
9. Distribution
Distributionofofthe
theincrease
increaseininsoil
soilmoisture
moistureatat
the mizuna
the greenhouse
mizuna after
greenhouse irrigation
after on on
irrigation 29
September 2015.
29 September 2015.

Based on the performance evaluations, it seems difficult with the developed sensor to detect
Based on the performance evaluations, it seems difficult with the developed sensor to detect
precise differences in soil moisture, e.g., less than 0.05 m33·m−−33, at different locations without further
precise differences in soil moisture, e.g., less than 0.05 m ·m , at different locations without further
improvements to reduce individual differences among sensors. However, the developed sensor can
improvements to reduce individual differences among sensors. However, the developed sensor can
detect relatively large differences (>0.05 m33·m− −3) in a field or among fields, and thus can be utilized to
detect relatively large differences (>0.05 m ·m 3 ) in a field or among fields, and thus can be utilized to
make decisions on irrigation volumes.
make decisions on irrigation volumes.
3.3. Possible Improvements of the Developed Sensor
The developed sensor captured dynamic changes in soil moisture at each depth, but some
opportunities for improvement were found though the two experiments.
The most serious problem of the developed sensor was that the sensor was sensitive to the
Sensors 2016, 16, 1292 11 of 14

3.3. Possible Improvements of the Developed Sensor

The developed sensor captured dynamic changes in soil moisture at each depth, but some
opportunities for improvement were found though the two experiments.
The most serious problem of the developed sensor was that the sensor was sensitive to the contact
between the sensor capacitors and the soil. The measured soil moisture kept increasing in the beginning
of the measurement as contact between the soil and sensors improved due to soil settling, and it took a
few weeks for the sensors to show reasonable soil moisture. The differences among individual sensors
stem from the variations in contact between the sensor capacitors and the soil. It was observed that the
soil moisture measurement suddenly decreased as a result of a human-induced change in the contact
between sensors and soils. This problem can be addressed by increasing the frequency of the capacitive
touch IC as much as possible. In this study, we used approximately 62 kHz with the MPR121, but this
value can be increased up to, for example, 10 MHz by using other capacitive touch ICs that allow us to
control the measurement frequency. This will expand the sampling volume of the sensor, which will
mitigate the dependence on contact condition between capacitor and soil. The problem may also be
solved by changing the sensor design. Because of the current sensor design, the sensor was inserted by
digging a hole, placing the sensor in the hole, and filling the gap between the sensor and hole with soil.
This installation procedure is the main cause for the sensor needing a few weeks before starting stable
soil moisture measurements. A solution would be to develop sensors designed such that they can
be hammered into the soil, e.g., sensors with spiky or plate-like shapes. Moreover, as the developed
sensors have a sensing part (the film substrate rolled on a PVC pipe), a measurement and data-logging
part (the box with a microcomputer and IC), and a communication unit, they require a relatively large
space for field installation. This caused workers in the field to accidentally touch the sensors. Therefore,
we are developing a new sensor design in which all parts are minimized and combined into a single
package. This also will reduce the sensor price and may enable easy waterproofing.
The contact between the film substrate and the PET heat-shrinking film as well as the contact
between sensor and soil is considered to be a cause of differences among individual sensors. The PET
heat-shrinking film was used to isolate the capacitor from the soil and to hold the substrate on the PVC
pipes. Each sensor had a minor difference in the contact between the two films, which was sufficiently
small compared to the sensor value. In order to avoid this contact problem, resist ink can be used to
isolate the capacitors, instead of the PET heat-shrinking film. It will contribute to making differences
between individual sensors negligible.
To address a different issue, a data re-sending function was put in place during this study.
With regard to the sudden drops in the measured capacitance, alternatives to the low-cost capacitive
touch IC are currently being tested. We have paid less attention to the waterproofing of the
communication unit until two of the nine sensors in the Ibaraki greenhouse field malfunctioned
due to the irrigation. After that, we have waterproofed the communication unit by taping the slight
space between the box and the cap and we have not seen any sensor malfunction due to wetting of the
communication unit. Also the communication unit is included inside the waterproofed main box in
the new sensor design we are developing, which will render this issue moot.
Another issue we have to investigate in the future is the influence of different types of soils on
the relationship between sensor output, i.e., capacitance, and VWC. In this study, the soils in the two
fields were similar and, thus, the calibration was only performed once. However, it is known that the
relationship between the sensor output and VWC often depends on soil types. Most of the currently
commercialized soil moisture sensors are provided with several calibration equations to adapt to a
variety of soils, enabling users to choose the appropriate one based on the target soil. The developed
sensor must be tested in several different soil types and the impact must be evaluated for expanding
the sensor usability.
Sensors 2016, 16, 1292 12 of 14

4. Conclusions
In this study, a low-cost soil moisture sensor using capacitors on a film substrate and a capacitive
touch IC was developed, and the performance of the sensor was evaluated in field experiments. Based
on the results from the experiments, the developed sensor could capture the dynamic change in
soil moisture at 10, 20, and 30 cm depth with some time required after sensor installation for the
contact between capacitors to settle down. The measured soil moistures showed the influence of
individual sensor differences, and this influence masked minor differences of less than 0.05 m3 ·m−3 in
soil moisture at different locations. However, the developed sensor could detect large differences of
more than 0.05 m3 ·m−3 and dynamic changes in soil moisture. Possible improvements to the design of
the developed sensor in future were also discussed based on the two field experiments.
The initial cost of a single developed sensor, i.e., the single slave node in Figure 3, was
approximately 200 U.S. dollars, and the sale price could be around 300 U.S. dollars. The sale price for
the master node is also around 300 U.S. dollars. These costs are much lower than those of soil moisture
sensors currently commercialized for research use. With further improvements and mass production,
the sensor cost could be reduced even more. This low-cost sensor will therefore be more affordable to
farmers as it requires low financial investment, and will thus allow expansion of precision agriculture.

Acknowledgments: This work was supported by grants from a project of the NARO Bio-oriented Technology
Research Advancement Institution (Integration Research for Agriculture and Interdisciplinary Fields). The authors
thank Yosuke Tamatsukuri, Union Farm Co., Ltd., for his support of this project.
Author Contributions: Kazuhiro Nishioka, Masaru Mizoguchi, and Yoshihiro Kawahara conceived and
designed the experiments; Yuki Kojima, Ryo Shigeta, Naoya Miyamoto, and Yasutomo Shirahama performed
the experiments; Yuki Kojima, Ryo Shigeta, Naoya Miyamoto, and Yasutomo Shirahama analyzed the data;
Masaru Mizoguchi and Yoshihiro Kawahara contributed with reagents/materials/analysis tools; Yuki Kojima
wrote the paper.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Fares, A.; Alva, A.K. Evaluation of capacitance probes for optimal irrigation of citrus through soil moisture
monitoring in an entisol profile. Irrig. Sci. 2000, 19, 57–64. [CrossRef]
2. Vellidis, G.; Tucker, M.; Perry, C.; Kvien, C.; Bednarz, C. A real-time wireless smart sensor array for scheduling
irrigation. Comput. Eectron. Agric. 2008, 61, 44–50. [CrossRef]
3. Kim, Y.; Evan, R.G.; Iverson, W.M. Remote sensing and control of an irrigation system using a distributed
wireless sensor network. IEEE Trans. Instrum. Meas. 2008, 57, 1379–1387.
4. Bouyoucus, G.J.; Mick, A.H. Improvements in the plaster of paris absorption block electrical resistance
method for measuring soil moisture under field conditions. Soil Sci. 1946, 63, 455–465. [CrossRef]
5. Spaans, E.J.A.; Baker, J.M. Calibration of Watermark soil moisture sensors for soil matric potential and
temperature. Plant Soil 1992, 143, 213–217. [CrossRef]
6. Noborio, K.; Horton, R.; Tan, C.S. Time domain reflectometry probe for simultaneous measurement of soil
matric potential and water content. Soil Sci. Soc. Am. J. 1999, 63, 1500–1505. [CrossRef]
7. Bouyoucus, G.J.; Mick, A.H. A fabric absorption unit for continuous measurement of soil moisture in the
field. Soil Sci. 1948, 66, 217–232. [CrossRef]
8. Gardner, W.; Kirkham, D. Determination of soil moisture by neutron scattering. Soil Sci. 1952, 73, 391–401.
[CrossRef]
9. Van Bavel, C.H.M.; Underwood, N.; Swanson, R.W. Soil moisture measurement by neutron moderation.
Soil Sci. 1956, 82, 29–41. [CrossRef]
10. Long, I.F.; French, B.K. Measurement of soil moisture in the field by neutron moderation. Eur. J. Soil Sci.
1967, 18, 149–166. [CrossRef]
11. Topp, G.C.; Davis, J.L.; Annan, A.P. Electromagnetic determination of soil water content: Measurements in
coaxial transmission lines. Water Resour. Res. 1980, 16, 574–582. [CrossRef]
12. Noborio, K. Measurement of soil water content and electrical conductivity by time domain reflectometry:
A review. Comput. Electron. Agric. 2001, 31, 213–237. [CrossRef]
Sensors 2016, 16, 1292 13 of 14

13. Dalton, F.N.; Herkelrath, W.N.; Rawlins, D.S.; Rhoades, J.D. Time-domain reflectometry: Simultaneous
measurement of soil water content and electrical conductivity with a single probe. Science 1984, 224, 989–990.
[CrossRef] [PubMed]
14. Topp, G.C.; Davis, J.L.; Annan, A.P. Electromagnetic determination of soil water content using TDR: II.
Evaluation of installation and configuration of parallel transmission lines. Soil Sci. Soc. Am. J. 1982, 46,
678–684. [CrossRef]
15. Nissen, H.H.; Moldrup, P.; Henriksen, K. High-resolution time domain reflectometry coil probe for measuring
soil water content. Soil Sci. Soc. Am. J. 1998, 62, 1203–1211. [CrossRef]
16. Bell, J.P.; Dean, T.J.; Hodnett, M.G. Soil moisture measurement by an improved capacitance technique, part II.
Field techniques, evaluation and calibration. J. Hydrol. 1987, 93, 79–90. [CrossRef]
17. Dean, T.J.; Bell, J.P.; Baty, A.J.B. Soil moisture measurement by an improved capacitance technique, part I.
Sensor design and performance. J. Hydrol. 1987, 93, 67–78. [CrossRef]
18. Tomer, M.D.; Anderson, J.L. Field evaluation of a soil water-capacitance probe in a field sand. Soil Sci. 1995,
159, 90–98. [CrossRef]
19. Nadler, A.; Lapid, Y. An improved capacitance sensor for in situ monitoring of soil moisture. Aust. J. Soil Res.
1996, 34, 361–368. [CrossRef]
20. Bogena, H.R.; Huisman, J.A.; Oberdörster, C.; Vereecken, H. Evaluation of a low-cost soil water content
sensor for wireless network applications. J. Hydrol. 2007, 344, 32–42. [CrossRef]
21. Seyfried, M.S.; Grant, L.E. Temperature effects on soil dielectric properties measured at 50 MHz. Vadose Zone J.
2007, 6, 759–765. [CrossRef]
22. Evett, S.R.; Tolk, J.A.; Howell, T.A. Soil profile water content determination: Sensor accuracy, axial response,
calibration, temperature dependence, and precision. Vadose Zone J. 2006, 5, 894–907. [CrossRef]
23. Evett, S.R.; Schwartz, R.C.; Tolk, J.A.; Howell, T.A. Soil profile water content determination: Spatiotemporal
variability of electromagnetic and neutron probe sensors in access tubes. Vadose Zone J. 2009, 8, 926–941.
[CrossRef]
24. Mazahrih, N.Th.; Katbeh-Bader, N.; Evett, S.R.; Ayars, J.E.; Trout, T.J. Field calibration accuracy and utility of
four down-hole water content sensors. Vadose Zone J. 2008, 7, 992–1000. [CrossRef]
25. Kelleners, T.J.; Robinson, D.A.; Shouse, P.J.; Ayars, J.E.; Skaggs, T.H. Frequency dependence of the complex
permittivity and its impact on dielectric sensor calibration in soils. Soil Sci. Soc. Am. J. 2005, 69, 67–76.
26. Paltineanu, I.C.; Starr, J.L. Real-time soil water dynamics using multisensory capacitance probes: Laboratory
calibration. Soil Sci. Soc. Am. J. 1997, 61, 1576–1585. [CrossRef]
27. Kizito, F.; Campbell, C.S.; Campbell, G.S.; Cobos, D.R.; Teare, B.L.; Carter, B.; Hopmans, J.W. Frequency
electrical conductivity and temperature analysis of a low-cost capacitance soil moisture sensor. J. Hydrol.
2008, 352, 367–378. [CrossRef]
28. Padilla, F.M.; Pugnaire, F.I. Rooting depth and soil moisture control Mediterranean woody seedling survival
during drought. Funct. Ecol. 2007, 21, 489–495. [CrossRef]
29. Bales, R.C.; Hopmans, J.W.; O’Geen, A.T.; Meadows, M.; Hartsough, P.C.; Kirchner, P.; Hunsaker, G.T.;
Beaudette, D. Soil moisture response to snowmelt and rainfall in Sierra Nevada mixed-conifer forest.
Vadose Zone J. 2011, 10, 786–799. [CrossRef]
30. Kojima, Y.; Mitsuishi, S.; Mizoguchi, M. Effectiveness of ICT field monitoring linking field images with
soil and environmental data: Case of a cold upland cabbage field. J. Jpn Soc. Soil Phys. 2015, 131, 5–13.
(In Japanese)
31. Shibusawa, S. Prevision farming Japan model. Agric. Inform. Res. 2003, 12, 125–133.
32. Kawahara, Y.; Hodges, S.; Cook, B.S.; Zhang, C.; Abowd, G.D. Instant inkjet curcuits: Lab-based inkjet
printing to support rapid prototyping of UbiComp devices. In Proceedings of the ACM Ubicomp 2013,
Zurich, Switzerland, 8–12 September 2013; pp. 363–372.
33. Kawahara, Y.; Lee, H.; Tentzeris, M.M. SenSprout: Inkjet-printed soil moisture and leaf wetness sensor.
In Proceedings of the ACM Ubicomp 2012, Pittsburgh, PA, USA, 5–8 September 2012; pp. 545–545.
34. Shirahama, Y.; Shigeta, R.; Kojima, Y.; Nishioka, K.; Kawahara, Y.; Asami, T. Implementation of wide range
soil moisture profile probe by coplanar plate capacitor on film substrate. In Proceedings of the 2015 IEEE
Sensors, Busan, Korea, 1–4 November 2015; pp. 1758–1761.
Sensors 2016, 16, 1292 14 of 14

35. NXP Semiconductors. MPR121, Proximity Capacitive Touch Sensor Controller-Data Sheet. NXP
Semiconductors: Eindhoven, The Netherlands, 2013. Available online: https://www.nxp.com/files/sensors/
doc/data_sheet/MPR084.pdf (accessed on 10 August 2016).
36. Delin, K.A.; Jackson, S.P.; Johnson, D.W.; Burleigh, S.C.; Woodrow, R.R.; McAuley, J.M.; Dohm, J.M.; Ip, F.;
Ferreé, T.P.A.; Rucker, D.F.; et al. Environmental studies with the sensor web: Principles and Practice. Sensors
2005, 5, 103–117. [CrossRef]
37. Pierce, F.J.; Elliott, T.V. Regional and on-farm wireless sensor network for agricultural systems in Eastern
Washington. Comput. Electron. Agric. 2008, 61, 32–43. [CrossRef]
38. Cobos, D. Measurement Volume of Decagon Volumetric Water Content Sensors. Decagon Devices,
Inc.: Pullman, WA, USA, 2015. Available online: http://manuals.decagon.com/Application%20Notes/
14955VWC%20Sensor%20Measurement%20Volumes_Web.pdf (accessed on 10 August 2016).

© 2016 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/).