This article has been accepted for publication in a future issue of this journal, but has not been
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2021.3069894, IEEE Internet of
Things Journal
An IoT-enabled Portable Water Quality
Monitoring System with MWCNT/PDMS
Multifunctional Sensor for Agricultural
Applications
Fowzia Akhter, Student Member, IEEE, Hasin R. Siddiquei, Md Eshrat E. Alahi, Member, IEEE,
Krishanthi Jayasundera, and S. C. Mukhopadhyay, Fellow, IEEE
 crucial roles in plant growth. Nitrate-N enters the ecosystem in
Abstract— The need to develop a low-power, low-cost nitrate,
phosphate, and pH sensor and sensing system is essential for
monitoring water quality in real-time. A novel interdigital sensor
has been fabricated and characterized for temperature, nitrate,
phosphate, and pH detection in water. The sensor is fabricated
using the 3D printing technique where electrodes are formed using
Multi-Walled Carbon Nanotubes (MWCNTs), and the substrate is
developed using Polydimethylsiloxane (PDMS). The sensor is
characterized by Electrochemical Impedance Spectroscopy (EIS)
to determine various temperatures, pH levels, nitrate, and
phosphate concentrations. Experimental outcomes prove that the
developed sensor can distinguish nitrate and phosphate
concentrations ranging from 0.1 ppm ~ 30 ppm, pH values from
1.71 ~ 12.59, temperature from 0~45℃. The sensitivity for
temperature, nitrate, phosphate, and pH level of the sensor are
1.1974Ω/℃, 1.9396Ω/ppm, 0.8839Ω/ppm and 1.0295Ω,
respectively. A location-independent portable smart sensing
system with LoRa connectivity is also developed to surveil water
quality and get feedback from the experts. A machine learning
algorithm trains the Arduino-based system and determines
temperature, nitrate and phosphate concentrations, and pH level
in real water samples. All the outcomes are compared with the
standard method for validation. The sensor and the sensing
system's performances are highly stable, reliable, and repeatable
to be a part of a smart sensing network for continuous water
quality monitoring.
Index Terms— EIS, Impedance, MWCNT, PDMS, Nitrate,
Phosphate, pH.
I. INTRODUCTION
W ATERcontamination is a major environmental issue
worldwide, and continuous monitoring of containing
ions is essential for keeping the drinking water, freshwater, and
groundwater safe. Among different ions, nitrate (NO3-),
phosphate (PO₄³⁻), and pH, i.e., hydrogen (H+) ions, play
Manuscript submitted December 1, 2020. (Corresponding author: Fowzia
Akhter.)
Fowzia Akhter, Hasin R. Siddiquei and S.C. Mukhopadhyay, are with Dept.
of Engineering, Macquarie University, Sydney, NSW 2109, Australia. (e-mail:
fowzia.2k3@gmail.com;engr_hasin96@yahoo.com;subhas.mukhopadhyay@
mq.edu.au).
2327-4662 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Prince Edward Island. Downloaded on May 15,2021 at 16:19:21 UTC from IEEE Xplore. Restrictions apply.
1
living and dead organisms and contributes to plant and animal
tissue production [1-2]. Although it helps animals and plants
synthesize protein, it may pollute rivers, streams, lakes, and
wetlands when present in excess amounts in the environment
[3]. The enrichment of nitrate-N of surface and groundwater is
a critical issue for animal farming in an agricultural country like
Australia [4]. The deposited urine in the dairy farm contains a
high amount of nitrate, and there is a severe concern of nitrate
leaching to the surface and groundwater. High concentrations
of nitrate-N can cause methemoglobinemia (blue baby
syndrome) if the infants drink water with highly concentrated
nitrate-N regularly [5]. As per the WHO, the maximum
permissible limit of nitrate is 11.3 ppm in the drinking water
[6].
Phosphorus is another essential mineral to generate usable
energy from the sunlight. It helps cells to grow and reproduce.
Therefore, a small amount of phosphorus loss from soil may
reduce algae and freshwater weeds growth [7-8]. Phosphorus
also damages the environment when added to the lake in many
amounts due to inconsiderable human activities, including
urban sewage, street runoff, and rural home septic tanks [9].
Suppose the phosphorous concentration regarding phosphates
exists in lakes or reservoirs more than 1 ppm. In that case, it
might affect water bodies' natural balance [10].
The pH level is a crucial factor deciding corrosivity of water
[11]. It is the measurement of the acidity or alkalinity of water
by determining hydrogen ions (H+). Water with a pH level
below 7.0 is acid and a pH level above 7.0 is considered bases
[12-13]. Rapid changes in the pH level of water may have a
significant impact on many organisms. Various aquatic animals
are adapted to living in certain water pH. Hence, they may die
if the water pH value varies slightly [14]. Drinking water
having a pH level of more than 8.5 may cause aesthetic
Md Eshrat E. Alahi is with Shenzhen Institute of Advanced Technology,
Chinese Academy of Sciences, Shenzhen 518055, China (e-mail:
ealahi@yahoo.com).
Krishanthi Jayasundera is with University Technology Sydney, 15 Broadway,
Ultimo NSW 2007, Australia, <Krishanthi.Jayasundera@uts.edu.au>
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2021.3069894, IEEE Internet of
Things Journal
2

problems [15]. Hence, it is necessary to have the best nitrate,
phosphate, and pH management in water used for farming for
both economic and environmental reasons.
There are numerous methods have been advised in analytical
chemistry to determine the water contaminants, such as High-
performance liquid chromatography (HPLC) [16], UV-
spectroscopy [17], chromatography [18], and colorimetric
methods [19] to detect nitrate and phosphate from water.
Although these expensive systems provide highly accurate
results, they are time-consuming, complex to operate, and need
daily care such as cleaning, storage, and maintenance [20].
Thus, it is essential to develop low-cost, low-power sensing
devices for detecting nitrate, phosphate, and pH accurately in
different water bodies to ensure a sustainable world to live.
Numerous researches have been done for determining the
contaminants of the water, such as nitrate, phosphate, chloride,
magnesium, calcium, and pH [21- 23]. However, it still needs
high-sensitivity, low-power, low-cost, and easily operable
sensors applicable to any environmental condition. Internet of
Things (IoT) enabled smart devices can also be used to monitor
environments and water quality, making our lives accessible to
the digital world [24]. The inclusion of communication
protocols enables data transfer from the sensor node to the
server to further process and predict future trends [25].
There are various materials available for developing the
substrates of the flexible sensors, including Polyethylene
terephthalate (PET) [26], Polyimide [27], and
Polydimethylsiloxane (PDMS) [28]. Among them, PDMS has
been chosen in this research as it is biocompatible,
hydrophobic, cost-effective. The interfacial bonds formed with
the sensors' electrodes and PDMS are outstanding [29]. Among
the conductive materials (Carbon Nanotubes (CNTs),
Aluminum, Gold, and Graphene), Multi-Walled Carbon
Nanotubes (MWCNTs) are utilized for developing electrodes
because it has better bonding capability and higher electrical
conductivity with PDMS [30]. This research discusses the
fabrication, and application of a novel MWCNTs/PDMS sensor
for nitrate, phosphate, and pH detection in water bodies. The
mobility and behaviour of the nutrient ions in the water change
when temperature changes [31]. Therefore, the sensitivity of the
sensor towards temperature is also studied. It also explains a
portable sensing system's design and development process to
monitor the farm's water quality. The portable system has the
Internet of Things (IoT) capability, which will help check the
gradients of the water's nutrients and pH level. Irrespective of
the farm's time and place, the farmer will be getting feedback
from the experts if any sudden environmental variations occur
inside the farm.
This article is divided into four sections. A brief introduction
of the proposed research is discussed in section I; the materials
and methods, including the fabrication process, working
principle, and experimental setup, design of the portable
system, are discussed in section II. The experimental outcomes
and development of the calibration standards are included in
section III. The research conclusion has been provided in
Section IV of the paper.
II. MATERIALS AND METHODS
This section represents the fabrication procedure and
experimental procedures for nitrate, phosphate concentrations,
and pH level determination using the developed sensor in water
bodies.
A. Fabrication Procedure
The master mould is designed using Autodesk Fusion 360
software and printed with Poly Lactic Acid (PLA) filament
using Dreamer 3D printer. At first, conductive ink is prepared,
mixing the MWCNTs powder and PDMS at a ratio of 10:2. The
optimum ratio is obtained by analyzing the variation of resultant
conductivity with the change in concentrations. The analysis
shows that the electrodes' resultant conductivity changes with
the variation of MWCNTs and PDMS ratio to a certain point.
After that, it reaches its saturation point. Therefore, a 10:2 ratio
is finalized for developing electrodes of the sensor. Initially, the
conductive ink is cast on the moulds. This is followed by curing
the moulds at 70˚C for an hour to solidify the nanocomposites.
After that, PDMS is poured on the mould to form the substrate.
A casting knife is used for adjusting the height of PDMS to
1000 µm. The samples are then put inside a desiccator until the
trapped air bubbles are removed. Finally, the sample is heated
at 70o C for three hours for solidification. Once the curing is
done, the samples are peeled off, and experiments are
conducted after washing. The sensor consists of six pairs of
interdigitated electrodes. The length of each electrode is 30 mm,
whereas the width is 1.5 mm. There is a 3 mm gap between two
successive electrodes. A good quality filament is a must to
ensure a smooth texture of the printed mould and making the
sensor reproducible. PDMS and CNT powder should be mixed
properly to ensure the sensor's conductivity. It is also essential
to provide adequate attention to remove the trapped bubbles by
desiccation. The schematic diagram of fabrication process is
shown in Fig.1.
Fig. 1. Schematic diagram of the sensor fabrication process.

2327-4662 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Prince Edward Island. Downloaded on May 15,2021 at 16:19:21 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2021.3069894, IEEE Internet of
Things Journal
3

The operating mechanism of the proposed sensor and parallel water for consecutive measurement. It is then dried for ten
plate capacitor is similar. When a low-amplitude alternating minutes for successive measurement. Finally, the experimental
voltage is applied between the excitation and sensing results are analyzed for developing calibration standards of
electrodes, an electric field initiates between them. The electric nitrate, phosphate, and pH from water samples.
field bulges out from the sensing electrode to the excitation
electrode due to the planar structure. This electric field changes
as soon as the material under test (MUT) is kept close to this
generated electric field. As a result, the values of the
impedances change. The MUT characteristics are analyzed in
this research by studying the variation of the electric field
properties. The sensor's operating principle are shown in Fig. 2.

Fig.2. Sensor’s working principle. Fig. 3. Experimental setup for nitrate and phosphate detection.

B. Experimental Process C. Design and development of the portable sensing system

The experiments for detecting nitrate and phosphate solution A portable system is designed using Autodesk Fusion 360
using the MWCNTs/PDMS sensor are conducted at ambient software and printed using a 3D printer. The proposed system
temperature and humidity. The experimental setup for taking consists of two parts: a box and a drawer. The box is the outer
the measurements using the developed sensor patch is shown in shell that holds the Liquid Crystal Display (LCD), and the
Fig.3. The measurement and instrumentation errors are avoided drawer is used to accommodate all the electronic devices. The
by taking necessary precautions before conducting the proposed system is composed of an AD5933 impedance
experiments, such as cleaning thoroughly and drying the analyzer [32], Nitrate, Phosphate, and pH sensor, LCD [33] for
beakers in the oven before preparing the solution so that there showing the measured data locally, Arduino Uno
are no other molecules and drop of water inside the beaker. Microcontroller [34] for data processing, LoRa shield [35] for
Various concentrations (0.1, 1, 10, 20, and 30 ppm) of nitrate sending data wirelessly to a cloud server named ThingSpeak
and phosphate solutions are prepared before starting the [36], charger shield [37] for charging the battery, and battery
experiment. The sample solutions are prepared using Potassium [38] to provide continuous energy. Fig. 4 shows the 3D design
nitrate (KN03) and Sodium phosphate (Na3PO4) as the solutes of the portable system, Fig.5 shows the connection diagram,
and deionized water (pH: 6.71, Resistance: 18.2 MΩ cm) as the and Fig.6 displays the final prototype.
solvent. Initially, 100 ppm solution is prepared and used as a
stock solution. Later, other solutions are prepared, applying the
serial dilution method. In addition to that, three acidic (1.71,
3.47, and 5.35) and basic (8.5, 10.49, and 12.59) solutions are
prepared, mixing Sulfuric acid (H2SO4) and Sodium Hydroxide
(NaOH) solutions respectively to deionized water.
The built-in open and short calibration tests are carried out
initially on the HIOKI IM 3536 impedance analyzer for
nullifying any stray capacitance due to connecting leads. After
that, it is tested in slow mode for achieving an error rate of less
than 0.05% as per manufacturer’s recommendation. A gold
contact clamp pin connector is used to establish the connection
between the HIOKI IM 3536 impedance analyzer and the
sensor. It is verified that sensor patch and the impedance
analyzer's Kevin probes are tightly connected. A laptop is Fig.4. 3D design of the proposed system.
attached to the impedance analyzer via USB cable for data
collection. While experimenting, it is confirmed that the sensor
patch's sensing area is appropriately dipped inside the solution.
The sensor patches' responses for resistance, reactance, and
impedance are obtained for the 10 Hz to 100 kHz frequency
range. The sensors patch is washed thoroughly using deionized

2327-4662 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Prince Edward Island. Downloaded on May 15,2021 at 16:19:21 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2021.3069894, IEEE Internet of
Things Journal
4

TABLE I
TRAINING DATASET MATRIX
Nitrate Phosphate T pH Z R ϴ
0.1 0.1 5 1.71 Z1 R1 ϴ1
0.1 0.1 5 3.47 Z2 R2 ϴ2
- - - - - - -
- - - - - - -
0.1 0.1 10 1.71 Z5 R5 ϴ5
0.1 0.1 10 3.47 Z6 R6 ϴ6
- - - - - - -
- - - - - - -
0.1 1 5 1.71 Z17 R17 ϴ17
0.1 1 5 3.47 Z18 R18 ϴ18
- - - - - - -
- - - - - - -
1 0.1 5 1.71 Z65 R65 ϴ65
1 0.1 5 3.47 Z66 R66 ϴ66
Fig.5. Connection diagram of the proposed system.
- - - - - - -
- - - - - - -
10 10 20 10.49 Zn Rn ϴn (n=256)

The sensor is operated at 1.3 kHz during the lab experiment

for sensing three parameters: impedance, resistance, and
reactance. In this way, 256 sets of solutions are tested by the
sensor. All these collected data are stored in a float array
variable matrix into Arduino IDE (Integrated Development
Environment) to train the system based on the KNN algorithm.
This program uses Arduino_KNN.h library function, which
classifies a new entry based on the majority of K-Nearest
Neighbor category. It is one of the popular algorithms for
pattern recognition and used by many researchers for sensor
calibration and data classification [39,40]. The data read by the
sensor is an unclassified data. The data processing algorithm
calculates the nearest distance between the new data and all the
data in the trainer matrix using the Euclidean distance's formula
[41].

𝑑(𝑧, 𝜃) = 𝑑(𝜃, 𝑧) = √∑𝑛𝑖=1(𝑧𝑖 − 𝜃𝑖 )2 (1)

The algorithm then optimizes the 'k' parameter from the

Fig.6. Final Prototype used for monitoring water quality.
D. Data Processing algorithm and validation
In real-life scenarios, water has many contaminants, so the
sensor's impedance depends on the combined effect. Therefore,
it is essential to record the sensor's performance where samples
are mixed and create a database to estimate the individual
nutrient amount in their compound solution. This database is
used in a machine learning algorithm (KNN) to train the
microcontroller to find the individual nutrient's nearest
probable concentration. For this purpose, a 7x256 matrix has
been prepared from the laboratory-made solutions of the
various proportions of nitrate and phosphate (0.1, 1,5 and 10
ppm) with variable pH (1.71, 3.47, 6.8 and 10.49) at different
temperatures (5, 10, 15, and 20℃). This matrix uses 2692 bytes
(8%) of the Arduino Uno total storage space of 32256 bytes.
Table I shows the training dataset matrix.
known reading and training matrix, classifies the shortest
distance group of rows of the matrix, and calculates near-
matched probable result using the mean deviation of those
distances. Fig.7 shows the flow chart of the algorithm.
The client and server codes to run the system is written using
the Arduino IDE. The system starts and initializes AD5933 to
operate the sensor. The high-precision impedance analyzer,
AD5933 determines the impedance, resistance and reactance
values at each frequency for a solution. The gain of the
impedance analyzer has to be calibrated before starting any
measurement. These gain factors are utilized to determine the
sensor's total impedance, resistance, and reactance values at
each frequency for a solution. The system is operated at a
particular frequency, determined based on the next section's
experimental outcomes. After collecting the data from the
impedance analyzer, it is processed using a data processing
algorithm mentioned above to predict appropriate temperature,
pH level, nitrate, and phosphate concentration. After that, data
are displayed on the LCD to check locally and transmit to the
Thingspeak server through the TTN (The Things Network) [42]
gateway. The channel ID and Application Process Interface
(API) keys are given in the server code to transfer the data to

2327-4662 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Prince Edward Island. Downloaded on May 15,2021 at 16:19:21 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2021.3069894, IEEE Internet of
Things Journal
the allocated channel. During this Arduino program
development, the spreading factor (SF) is set to 7 for the fastest
data transmission. This Thingspeak server can be accessed from
any location. This will allow the agronomist to monitor the real-
time data for early detection of any ecological risk. Lora
Communication protocol is used for data transmission in the
proposed system as it consumes the least power and provides
broad coverage compared to other communication
protocols[43] such as Zigbee [44] and WiFi [45]. A
rechargeable battery is used rather than using the one-time
battery to be charged whenever unused. It will reduce the cost
of the system when used in the long-term.
Fig.7. Data processing algorithm of the proposed system.
2327-4662 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Prince Edward Island. Downloaded on May 15,2021 at 16:19:21 UTC from IEEE Xplore. Restrictions apply.
5
III. RESULTS AND DISCUSSIONS
This section represents the sensor responses towards
temperature, nitrate, phosphate concentrations, pH level and the
methods for developing the calibration standards to calculate
nitrate, phosphate concentrations and pH value in unknown
samples.
A. The Response towards Temperature, Nitrate, Phosphate,
and pH
This research aims to develop a portable nitrate, phosphate
sensing, and pH system applicable to indoor and outdoor
conditions. The temperature of the water varies significantly in
a practical scenario. As a result, the sensor’s performance
suffers due to temperature variations. Hence, the temperature
effect on the sensor's performance is studied initially. It is
carried out following the procedure based on prior research
[46]. The ions' mobility varies because of the change in
temperature, which causes variation in impedance. This
experiment's experimental setup is almost the same as Fig. 3,
which includes an IEC Hotplate (Cat no. 1920-001), a mercury
thermometer, HIOKI IM 3536 LCR meter, and a computer. The
deionized water temperature is varied from 5o C to 45o C using
IEC Hotplate, and the sensor's resistances, reactance, phase
angles and impedances are measured for 10 Hz ~ 100 kHz
frequency range. A mercury thermometer is dipped inside the
water for determining the actual temperature while taking this
measurement. The resistances of the sensor change
considerably in comparison to the change in reactance for
various temperatures. Therefore, the graph presenting the
change in resistance for various frequencies (Fig.8(a)) is
included in this article. The maximum sensitivity of the sensor
for temperature is calculated as 1.1974 Ω/℃. A frequency of
1.3 k Hz is selected as the operating frequency as it lies in the
sensitive region where all the temperatures are distinct. At this
frequency, the values of resistances are considered for finding
the correlation between the sensor's temperature and resistance.
Fig.8 (e) represents temperature and sensor's resistance
relationship. This curve is used for measuring temperature from
unknown water sample. The determination coefficient (R2) is
0.9902, which reveals a well-correlation between the
temperature and the sensor's resistance. Thus, temperature can
be measured using the following equation.
6×106 −𝑅𝑇
𝑇 (℃) = (2)
104539
Where, T shows the temperature determined in (oC), and RT is
the obtained resistance for a particular temperature. A constant,
αT = 104539 Ω/oC is the temperature coefficient.
The sensor is tested for different concentrations of nitrate,
phosphate, and pH solutions, and its responses are recorded.
The resistance and reactance changes are observed for various
concentrations (0.1 ppm to 30 ppm) of nitrate and phosphate as
well as for various values of pH (1.71 to 12.59). The sensor's
response towards nitrate, phosphate, and pH for different
frequencies are shown in Fig. 8 (b), Fig. 8 (c), and Fig. 8 (d).
Experimental outcomes proved that the sensor could
differentiate all the nitrate, phosphate, and pH samples. These
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2021.3069894, IEEE Internet of
Things Journal
6

changes are attributed to the double-layer capacitance (Cdl), sensitivity for nitrate, phosphate, and pH level of the sensor are
adsorption capacitance (Cad), and solution resistance (Rs). The calculated as 1.9396 Ω/ppm ,0.8839 Ω/ppm and 1.0295 Ω,
sensor shows considerable changes towards nitrate solutions for respectively.
the whole range of frequency (10 Hz ~ 100kHz); for phosphate, The highest value of R2 is achieved by applying the linear fit
the prominent response is found between 10 Hz to 10 kHz, for nitrate, phosphate, and pH. These regression analyses are
whereas, for pH, the significant changes are found between 10 simple and easy. Therefore, these have been chosen for
Hz to 300 kHz. It is observed from the experimental outcomes covering the whole range of nitrate, phosphate concentrations,
that the change in resistance is more prominent in comparison and pH levels.
to reactance towards the sample solutions. Thus, variation in
107 −𝑅𝑠𝑒𝑛𝑠𝑒
resistance concerning frequency towards different nitrate and For nitrate, 𝐶𝑛𝑖𝑡𝑟𝑎𝑡𝑒 = (3)
364104
phosphate concentrations and pH levels are included here. The 5×106 −𝑅𝑠𝑒𝑛𝑠𝑒
calibration curves (Fig. 8 (e)) are also achieved based on the For phosphate, 𝐶𝑝ℎ𝑜𝑠𝑝ℎ𝑎𝑡𝑒 = (4)
165918
resistance values against various nitrate, phosphate 7×106 −𝑅𝑠𝑒𝑛𝑠𝑒
For pH level, 𝑃𝐻_𝑙𝑒𝑣𝑒𝑙 = (5)
193248
concentrations, and pH values.
While obtaining the calibration curve for measuring the Where, Cnitrate (ppm), Cphosphate (ppm) refer to the actual nitrate
unknown concentrations, a particular frequency is chosen, and phosphate concentrations, PH_level is the pH value,
which is common for all the tested solutions of nitrate, whereas Rsense (Ω) shows the sensor's real impedance.
phosphate, and pH. Hence, the calibration standards are
B. Reproducibility and repeatability of the sensor
developed at 1.3 kHz to determine nitrate, phosphate
concentrations, and pH values with the sensor's resistance. The Five sensors are fabricated initially for testing their
reproducibility. Their responses towards 10 ppm phosphate

(a) (b)

(c) (d)

(e)
Fig. 8. Sensor's response towards various (a) temperatures, (b) nitrate, (c) phosphate, and (d) pH for a frequency sweep (10 Hz~100 kHz), and (e) The functional
relationship between real part of impedance and temperature, nitrate, phosphate, and pH.

2327-4662 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Prince Edward Island. Downloaded on May 15,2021 at 16:19:21 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2021.3069894, IEEE Internet of
Things Journal
7

concentrations are recorded. The outcomes are summarized in pH values in various water samples with more than 95%
Table II. It is found that all the sensors exhibit similar accuracy. The 5% measurement error is because of other
performances with a maximum RDS of around 1.5%. minerals' presence in the sample water. However, the accuracy
Therefore, it can be confirmed that the proposed fabrication can be further improved by preparing mixed samples with
technique promotes outstanding reproducibility. additional water quality parameter such as calcium,
magnesium, and dissolved oxygen and training the system with
TABLE II more data using a high memory-based microcontroller board.
TESTING REPRODUCIBILITY OF THE SENSOR

Actual Measured RSD

Nutrient Sensor concentration concentration by (%)
Number (ppm) the sensor
(ppm)
1 10.12
2 10.37
Phosphate 3 10 10.03 1.48
4 10.00
5 10.05

The sensor responses' towards differently concentrated

nitrate solutions are repeated three times to understand the
proposed sensors' repeatability at 1.3 kHz. Fig. 9 shows the
collected data for various nitrate solutions. Relative standard
deviation (RSD) is determined to know the sensor's responses
consistency. This signifies how the standard deviation of the
Fig. 10. Measurement of nitrate concentrations from unknown samples and
dataset changes compared to the mean. The lower the value of validation using the standard method.
RSD, the more precise the measured data are. The value of RSD
is found to be below 2%. Hence, it can be said that the proposed
sensors are highly repeatable.

Fig. 11. Measurement of phosphate concentrations from unknown samples and

validation using the standard method.
Fig. 9. Sensor's repeatability test towards various solutions concentrations
ranging from 0.1~30 ppm for nitrate.

C. Unknown Sample Measurement

After training the Arduino-based system, the system is used
to measure real samples from different lakes, rivers, creeks,
streams, and tap water of Sydney, Australia. As the
concentrations of nitrate and phosphate in the samples are not
too high for measuring in high-level detection range, some
nitrate and phosphate samples are added in sample numbers
3,4,5, and 6 to elevate the concentrations. The samples are used
as it is for pH value determination. The sensor's parameters
towards a particular water sample are collected, and KNN
algorithm is applied to find out nitrate, phosphate
concentrations, and pH level based on the trained samples. The
(c)
measured outcomes are compared with the standard lab-based Fig. 12. Measurement of pH from unknown samples and validation using the
method for validation. Fig. 10, Fig. 11, and Fig. 12 show the standard method.
sensor can determine nitrate and phosphate concentrations and

2327-4662 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Prince Edward Island. Downloaded on May 15,2021 at 16:19:21 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2021.3069894, IEEE Internet of
Things Journal
8

D. Data Storage into Thingspeak Cloud Server
The LoRa connectivity is included in the proposed system
so that the farmers can check the nutrients levels and transfer
the data in the Thingspeak channel for storing and sharing with
the experts. The agronomist can review the collected data
immediately in the IoT server. This will allow them to
understand the trend and advise the farmers accordingly from
any remote location. Therefore, the portable system is taken to
the Macquarie University Creek two times (morning and
afternoon) daily for a week to study its water quality, and data
are transferred to Thingspeak as shown in Fig. 13. The
concentration of nitrate and phosphate is below 3 ppm. The
temperature is varied from 10o~15oC. The concentrations of
nitrate and phosphate are not too high, interprets the quality of
water is good. The pH value varies from 8.0 ~8.5, which defines
the water is alkaline. As the pH level does not change
significantly, the organisms can live and grow adequately in the
creek water. This developed system will substantially impact on
taking decisions, making farms smart enough to adapt more
quickly to changing conditions. Remotely monitoring the farm
conditions and infrastructure helps saving time, labour, and
cost, enabling farmers to focus on other things.
cost, low-power portable system responds quickly, which is
within 1 second. Another essential feature of the proposed
sensor is reliability in determining pH level, nitrate, and
phosphate concentrations with RSD less than 3%. The design
and application of a portable smart sensing system for
monitoring water quality in real-time are also explained.
Inclusion of the K-NN machine learning algorithm enhanced
the applicability in real-sample with 95% accuracy. The
implementation cost of the system is about 150 US dollar which
includes the purchase of electronic consumables in small
numbers. However, the overall cost can be reduced
significantly when product is developed in large numbers. The
proposed sensor patch and the smart portable system are highly
potential to develop a complete and efficient water quality
surveillance system in real-time. This will enable early
detection of environmental hazards and provide time for an
effective countermeasure. Anyone can use this portable system
with minimal training. This is an ideal system for a regular
evaluation of contaminants level.
The low-cost sensor's performance will be enhanced further
by introducing the selectivity for nitrate, phosphate, pH, and
temperature as a coating or reliable sensing materials in future

(a) (b)

(c) (d)
o
Fig.13. (a) Temperature ( C), (b) Nitrate concentration (ppm), (c) Phosphate concentration (ppm), and pH level data transferred into the Thingspeak cloud server.

IV. CONCLUSION research. A high-performance microcontroller unit will

The fabrication and application of the MWCNTs/PDMS
sensor for smart agriculture applications are successfully
presented. The limit of detection of the proposed sensor for
nitrate, phosphate is 0.1 ppm, and for pH, it is 1.71, which is
highly promising for various agricultural applications. The low-
improve the KNN algorithm's performance with large data sets
and rapid processing. Additionally, a high-configuration
intelligent sensing system will be developed and installed in
remote locations, to track the water quality of lakes, rivers,
reservoirs, and creeks in Australia.

2327-4662 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Prince Edward Island. Downloaded on May 15,2021 at 16:19:21 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2021.3069894, IEEE Internet of
Things Journal
ACKNOWLEDGMENT
The authors would like to thank the School of Engineering,
Macquarie University, for providing the research facilities to
fabricate the sensor and perform necessary experiments.
REFERENCES
[1] Environmental impact of nitrogen and phosphorus fertilisers in high
rainfall areas (2020, Nov 1). [Online]. Available: https : // www. agric.
wa. gov. au/ high – rainfall - pastures / environmental – impact – nitrogen
– and -phosphorus-fertilisers-high-rainfall-areas.
[2] M. Masrie, M. S. A. Rosman, R. Sam, and Z. Janin, "Detection of
nitrogen, phosphorus, and potassium (NPK) nutrients of soil using optical
transducer", in Proc. IEEE International Conference on Smart
Instrumentation, Measurement and Applications (ICSIMA), Putrajaya,
Malaysia, 28-30 Nov., 2017.
[3] S. Harnsoongnoen, A. Wanthong, U. Charoen-In, and A. Siritaratiwat,
"Microwave sensor for nitrate and phosphate concentration sensing",
IEEE Sensors Journal, vol. 19, no. 8, pp.2950-2955 ,2019.
[4] Clifford S. Snyder, "Enhanced nitrogen fertilizer technologies support the
'4R' concept to optimize crop production and minimize environmental
losses", Soil Research- CSIRO Publishing, vol. 55, pp. 463–472, 2017.
[5] Linsha Ma, Liang Hu, Xiaoyu Feng, Songlin Wang, "Nitrate and Nitrite
in Health and Disease", Aging and Disease, vol. 9, no. 5, pp. 938-945,
2018.
[6] Md. Eshrat E. Alahi , N. Pereira-Ishak, S. C. Mukhopadhyay, and L.
Burkitt, "An Internet-of-Things enabled smart sensing system for nitrate
monitoring ", IEEE Internet of Things Journal, vol. 5, no. 6, pp. 4409-
4417, 2018.
[7] A. Nag, Md. Eshrat E. Alahi, Shilun Feng and S. C. Mukhopadhyay, "IoT-
based sensing system for phosphate detection using Graphite/PDMS
sensors", Sensors and Actuators A: Physical, Vol. 286, pp. 43–50, 2019.
[8] Md F. Kabir, Md T. Rahman, A. Gurung, and Q. Qiao, "Electrochemical
phosphate sensors using silver nanowires treated screen printed
electrodes", IEEE Sensors Journal, vol. 18, no. 9, pp. 3480-3485, 2018.
[9] C. S. Storer, Z. Coldrick, D. J. Tate, J. M. Donoghue, and B. Grieve,
"Towards phosphate detection in hydroponics using molecularly
imprinted polymer sensors", Sensors, vol. 18, no. 531, pp. 1-11, 2018.
[10] M. Sarwar, J. Leichner, G. M. Naja, and Chen-Zhong Li, "Smart-phone,
paper-based fluorescent sensor for ultra-low inorganic phosphate
detection in environmental samples", Microsystems & Nanoengineering,
vol.5, no. 56, pp.1-10, 2019.
[11] A. Saad, A. E. H. Benyamina, and A. Gamatié, "Water management in
agriculture: A survey on current challenges and technological Solutions",
IEEE Access, vol. 8, pp.38082-38097, 2020.
[12] N. Mandal, S. Mitra, and D. Bandyopadhyay, "Paper-sensors for point-
of-care monitoring of drinking water quality", IEEE Sensors Journal, vol.
19, no. 18, pp. 7936-7941, 2019.
[13] A. Kushagra , A. K. Pradhan, and D. Bazal, "pH-dependent
electrochemomechanical transition of hydrophobe-water interface", IEEE
Sensors Letters, vol. 3, no. 7, pp.1-4, 2019.
[14] S. Dudala, S. K. Dubey, and S. Goel, "Microfluidic soil nutrient detection
system:integrating nitrite, pH, and electrical conductivity detection",
IEEE Sensors Journal, vol. 20, no. 8, pp. 4504-4511, 2020.
[15] pH in Drinking-water (2020, Nov 25). [Online]. Available:
https://www.who.int/water_sanitation_health/dwq/chemicals/ph_revised
_2007_clean_version.pdf
[16] Mohau N. Moshoeshoe and Veronica Obuseng, "Simultaneous
determination of nitrate, nitrite and phosphate in environmental samples
by high performance liquid chromatography with UV detection", South
African Journal of Chemistry, vol. 71, pp. 79–85, 2018.
[17] Andreja Drolc and JanezVrtovšek, "Nitrate and nitrite nitrogen
determination in wastewater using on-line UV spectrometric method",
Bioresource Technology, vol.101, no.1, pp. 4228–4233, 2010.
[18] Md.Eshrat E.Alahi and S. C. Mukhopadhyay, "Detection methods of
nitrate in water: A review", Sensors and Actuators A: Physical, vol. 280,
pp. 210-221, 2018.
[19] V. Rérolle, D.Ruiz-Pino, M. Rafizadeh, S. Loucaides, S. Papadimitriou,
M. Mowlem, J. Chen, "Measuring pH in the Arctic Ocean:Colorimetric
method or SeaFET?", Methods in Oceanography, vol.17, pp.32-49, 2016.
2327-4662 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Prince Edward Island. Downloaded on May 15,2021 at 16:19:21 UTC from IEEE Xplore. Restrictions apply.
9
[20] A. Azmi, A. A. Azman, K. K. Kaman, S. Ibrahim, S. C. Mukhopadhyay,
S. W. Nawawi, and M. A. Md Yunus, "Performance of coating materials
on planar electromagnetic sensing array to detect water contamination",
IEEE Sensors Journal, vol. 17, no. 16, pp. 5244-5251, 2017.
[21] Md. Eshrat E. Alahi, Subhas Chandra Mukhopadhyay, and Lucy Burkitt,
"Imprinted polymer coated impedimetric nitrate sensor for real- time
water quality monitoring", Sensors and Actuators B: Chemical, vol. 259,
pp. 753–761, 2018.
[22] F. Akhter, A. Nag, Md. Eshrat E.Alahi, H. Liu, and S. C. Mukhopadhyay,
"Electrochemical detection of calcium and magnesium in water bodies",
Sensors and Actuators A: Physical, vol. 305, no. 111949, pp. 1–10, 2020.
[23] X. Wang, L. Sun, T. Wang, and Y. Shi, "Temperature-Compensated pH
Microelectrode Array Sensors Based on Copper-Oxide/Polyaniline
Modification", IEEE Sensors Journal, Vol. 20, No. 24, pp. 14598-14606,
2020.
[24] Fowzia Akhter, Sam Khadivizand, Hasin Reza Siddiquei, Md Eshrat E.
Alahi, and Subhas Mukhopadhyay, "IoT enabled intelligent sensor node
for a smart city: pedestrian counting and ambient monitoring", Sensors,
vol.3374, pp. 1-19, 2019.
[25] S. Pasika and S. T. Gandla, " Smart water quality monitoring system with
cost-effective using IoT", Heliyon, vol.6, 2020.
[26] Y. Wang, X. Wang, W. Lu, Q. Yuan, Y. Zheng, B. Yao, "A thin film
polyethylene terephthalate (PET) electrochemical sensor for detection of
glucose in sweat", Talanta, Vol. 198, pp. 86-92, 2019.
[27] Y. Qin, Q. Peng, Y. Ding, Z. Lin, C. Wang, Y. Li, et al., "Lightweight,
superelastic, and mechanically flexible graphene/polyimide
nanocomposite foam for strain sensor application", ACS nano, Vol. 9, pp.
8933-8941, 2015.
[28] T.H. da Costa and J.-W. Choi, "A flexible two dimensional force sensor
using PDMS nanocomposite", Microelectronic Engineering, vol. 174, pp.
64-69, 2017.
[29] Sam Ali, Dinesh Maddipatla, Binu Baby Narakathu, Amer A. Chlaihawi,
Sepehr Emamian, Farah Janabi, Bradley J. Bazuin, and Massood Z.
Atashbar, "Flexible capacitive pressure sensor based on PDMS substrate
and Ga–In liquid metal", IEEE Sensors Journal, vol. 19, no. 1, pp. 97-
104, 2019
[30] J. C. Dutta, and P. K. Sharma, "Fabrication, characterization and
electrochemical modeling of CNT based enzyme field effect
acetylcholine biosensor", IEEE Sensors Journal, vol. 18, no. 8,pp. 3090-
3097 , 2018.
[31] Md. R.R.Khan and S. W. Kang, "Fast, highly sensitive interdigitated
capacitor sensor to detect wide range of temperatures using graphene-
oxide-containing dielectric membrane", IEEE Sensors Journal, vol. 18,
no. 7,pp. 2667-2674, 2018.
[32] AD5933. Analogue Devices, 1 MSPS, 12-Bit Impedance Converter,
Network Analyzer (2020, Nov 5). [Online]. Available: https://
www.analog.com/media/en/technical-documentation/data-
sheets/AD5933. pdf
[33] LCD display. MC21605C6W-SPR-V2 (2020, Nov 5). [Online]. Available:
https://au.element14.com/midas/mc21605c6w-bnmlwi-v2/display-
alphanumeric-16x2-white/dp/2748649?gclid=CjwKCAjwgdX4BRB
_EiwAg8O8HZephXcGlbugLI3szR8P3zuoUJY6E7hT0WZaYXm9qEY
0OlADypQN9xoCPagQAvD_BwE&mckv=s_dc|pcrid|432287894061|p
kw||pmt||slid||product|2748649|pgrid|102205442804|ptaid|aud-
471864486199:pla-900921575218|&CMP=KNC-GAU-GEN-
SHOPPING-OPTOELECTRONICS-AND-DISPLAYS
[34] Arduino Uno Rev3 (2020, Nov 5). [Online]. Available:
https://store.arduino.cc/usa/arduino-uno-rev3
[35] Lora Shield for Arduino - Long Range Transceiver (2020, Nov 5).
[Online]. Available: https://www.iot-store.com.au/products/lora-shield-
for-arduino-long-range-transceiver
[36] Thingspeak (2020, Nov 5). [Online]. Available: https://thingspeak.com/
[37] Seeed studio solar charger shield v2.2(2020, July 6). [Online]. Available:
https://www.iot-store.com.au/products/seeed-studio-solar-charger-
shield-v2-2
[38] Spare Li-ion Battery to suit DC-1074 / DC-1093 (2020, Nov 8). [Online].
Available: https://www.jaycar.com.au/spare-li-ion-battery-to-suit-dc-
1074-dc-1093/p/DC1075
[39] N. Bokka, V. Selamneni, and P. Sahatiya, "A water destructible SnS2
QD/PVA film based transient multifunctional sensor and machine
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2021.3069894, IEEE Internet of
Things Journal
10

learning assisted stimulus identification for non-invasive personal care funds from the Chinese Academy of Sciences (CAS) and the
diagnostics", Material Advances, vol. 2818, 2020.
National Natural Science Foundation of China (NSFC), and the
[40] D. Adidrana and N.Surantha, "Hydroponic Nutrient Control System based
on Internet of Things and KNearest Neighbors", in Proc. IEEE
Ministry of Science and Technology (MOST) for the Foreign
International Conference on Computer, Control, Informatics and its Youth Project as a Principal Investigator (PI). He has research
Applications (IC3INA), Tangerang, Indonesia,, 23-24 Oct. 2019. interests in sensor design and fabrication, robust neural
[41] J. Vitola, F. Pozo, D.A. Tibaduiza ,and M. Anaya, "A sensor data fusion interface, and neural recording/stimulation devices for
system based on k-Nearest Neighbor pattern classification for structural biomedical applications.
health monitoring applications", Sensors, vol. 417, 2017.
[42] The Things Network (2020, Nov 22). [Online]. Available:
https://www.thethingsnetwork.org/ Dr. Krishanthi P. Jayasundera graduated
[43] J.Botero-Valencia, L. Castano-Londono , D. Marquez-Viloria, and M. from University of Peradeniya, Sri Lanka
Rico-Garcia, "Data reduction in a low-cost environmental monitoring with honours degree in Chemistry. She
system based on LoRa for WSN", IEEE Internet of Things Journal, vol.6, obtained her both Master and PhD in
no.2, pp.3024-3030, 2019.
Organic Chemistry from Kanazawa
[44] Zigbee Alliance(2020, Nov 10). [Online]. Available:
https://zigbeealliance.org/ University, Japan. She worked as post-
[45] Wi-Fi Alliance (2020, Nov 22). [Online]. Available: https://www.wi-
doctoral researcher at Massey University
fi.org/ nearly 14 years in New Zealand involving
[46] M.E.E. Alahi, L. Xie, S.C. Mukhopadhyay, and L. Burkitt, "A various projects focused on the chemical
temperature compensated smart nitrate-sensor for agricultural industry", synthesis of architecturally interesting molecules which have
IEEE Trans. on industrial electronics, vol.64, no.9, pp. 7333-7341, 2017. biological, environmental and /or medicinal
significance. Currently, she is working as Senior Researcher
Fowzia Akhter has completed B. Sc. in with University Technology Sydney, Australia. She specialises
Electrical and Electronic Engineering in Organic chemistry, Bio-sciences, sensitivity analysis using
from Khulna University of Engineering NMR, HPLC, SPR and so on. She has published over 30 papers
and Technology (KUET), Bangladesh in in different international journals and conference proceedings,
2008 and received her M.Sc in edited 5 books and 10 conference proceedings.
Communication Engineering from
International Islamic University Malaysia Dr. Subhas Chandra Mukhopadhyay is
(IIUM), Malaysia in 2012. Due to her working as a Professor of Mechanical/
excellent result during Masters, she received Rector's Award Electronics Engineering, Macquarie
which is one of the most prestigious awards for postgraduate University, Australia and is the Discipline
students of IIUM. Upon completion of M.Sc, she joined Linton Leader of the Mechatronics Engineering
University College, Malaysia as a lecturer. He served there Degree Programme. His fields of interest
before joining Macquarie University for her Ph.D. in 2017. Her include Smart Sensors and sensing
research interests include MEMS-based gas sensors, flexible technology, instrumentation techniques,
sensors, smart sensing systems, Internet of things, optics and wireless sensors and network (WSN),
photonics. Internet of Things (IoT), wearable sensors etc. He has
supervised over 40 postgraduate students and over 100 Honours
Hasin R. Siddiquei has completed B. Sc. students. He has examined over 60 postgraduate theses. He has
in Electrical and Electronic Engineering published over 400 papers in different international journals
from Rajshahi University of Engineering and conference proceedings, written 7 books and 42 book
and Technology (RUET), Bangladesh and chapters and edited seventeen conference proceedings. He has
received his M.Sc in Communication also edited thirty books with Springer-Verlag and twenty five
Engineering from International Islamic journal special issues. He has organized over 20 international
University Malaysia (IIUM), Malaysia. conferences as either General Chairs/co-chairs or Technical
During M.Sc, he developed a MEMS biosensor for detecting Programme Chair. He has delivered 353 presentations
DF-1 cells. After completion of M.Sc, he worked at including keynote, invited, tutorial and special lectures. He is
Infrastructure University Kuala Lumpur (IUKL), Malaysia as a a Fellow of IEEE (USA), a Fellow of IET (UK), a Fellow of
lecturer and Uttara University (UU), Bangladesh as a senior IETE (India), a Topical Editor of IEEE Sensors journal, and an
lecturer. Currently, he is working with Prof. Subhas associate editor of IEEE Transactions on Instrumentation and
Mukhopadyay as a visiting researcher at Macquarie University. Measurements. He is a Distinguished Lecturer of the IEEE
Sensors Council from 2017 to 2022. He is the Founding Chair
Md. Eshrat E. Alahi has received a Ph.D. of the IEEE Sensors Council New South Wales Chapter. More
in Engineering from Macquarie details can be available at
University, Sydney, Australia. He is http://web.science.mq.edu.au/directory/listing/person.htm?id=
currently working as a Post-Doctoral smukhopa
Research Fellow at Shenzhen Institutes of
Advanced Technology (SIAT), Chinese
Academy of Sciences. He has published
more than 35 research papers, including SCI journals,
conference proceedings, and book chapters. He has got research

2327-4662 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Prince Edward Island. Downloaded on May 15,2021 at 16:19:21 UTC from IEEE Xplore. Restrictions apply.