Article

Integrated Low-Cost Water Quality Monitoring System Based

on LoRa Network
Ioannis Georgantas 1 , Spyridon Mitropoulos 2, * , Stylianos Katsoulis 1 , Ioannis Chronis 1
and Ioannis Christakis 1

1 Department of Electrical and Electronics Engineering, University of West Attica, P. Ralli & Thivon 250,
12244 Egaleo, Greece; ee05866@uniwa.gr (I.G.); skatsoulis@uniwa.gr (S.K.); chroniio@uniwa.gr (I.C.);
jchr@uniwa.gr (I.C.)
2 Department of Surveying and Geoinformatics Engineering, University of West Attica, 28, Ag. Spyridonos Str.,
12243 Egaleo, Greece
* Correspondence: smitro@uniwa.gr

Abstract: Water quality is crucial for public health, especially in areas with water scarcity,
such as remote islands. Although reliable measuring devices are available, on-site manual
testing, such as in water tanks, is still necessary. In order to determine the quality of
drinking water, a set of measurements should be carried out automatically, at regular
intervals and without delay, thus ensuring the monitoring of its suitability. In this research
work, an integrated low-cost water quality monitoring system is presented. The proposed
system consists of the monitoring stations and an information system. Each monitoring
station includes a microcontroller and sensors, and is installed in the water reservoir, while
the information system is used to capture, store, and visualize the measurement data. Data
transmission is carried out over a long-range (LoRa) network, providing extensive coverage
for receiving data from Internet of Things (IoT) devices. Additionally, linear and non-linear
correction factors are proposed to improve the accuracy of measurements from low-cost
sensors, resulting in more reliable data. In this way, the end users, such as local authorities
or citizens, are given the possibility of immediate information on water quality via the web.

Keywords: water quality monitoring; LoRa network; TDS measurements; pH measurements;

low-cost IoT; distance communications; water quality
Academic Editor: Stefano Scanzio

Received: 2 February 2025

Revised: 16 February 2025
Accepted: 17 February 2025 1. Introduction
Published: 21 February 2025
Water is a vital ingredient for the existence and maintenance of life. Due to inap-
Citation: Georgantas, I.; Mitropoulos,
propriate management of water resources, severe water shortages have been observed
S.; Katsoulis, S.; Chronis, I.; Christakis,
in various residential areas, especially on islands during the summer months. Problems
I. Integrated Low-Cost Water Quality
Monitoring System Based on LoRa
with water may not only be related to luck or shortage, but also to the insufficiency of
Network. Electronics 2025, 14, 857. water of adequate quality. The importance of the availability of good quality water is
https://doi.org/10.3390/ demonstrated in the United Nations Sustainable Development Goals, where Goal No. 6
electronics14050857 is “Ensure availability and sustainable management of water and sanitation for all” [1].
Copyright: © 2025 by the authors. The provision of water pollution adversely affects health and poses a risk of illness and
Licensee MDPI, Basel, Switzerland. even death [2,3]. In line with the above, water quality should be continuously monitored.
This article is an open access article In recent years, the development of technology has led to the creation of both reliable,
distributed under the terms and
low-cost sensors and wireless telecommunications networks, which can support water
conditions of the Creative Commons
quality monitoring systems. A water quality monitoring system should measure a range of
Attribution (CC BY) license
(https://creativecommons.org/
parameters such as total dissolved solids (TDS), temperature, and pH in the water reservoir
licenses/by/4.0/). of a region.

Electronics 2025, 14, 857 https://doi.org/10.3390/electronics14050857

Electronics 2025, 14, 857 2 of 17

Several research groups have presented water quality measurement systems, such
as in [4], where a low-cost, real-time water quality measurement system consisting of an
Arduino microcontroller and sensors is presented. In work [5], using an embedded system
consisting of sensors based on IoT for testing water quality in a river, the water quality is
assessed by weighted learning models and, in the case of high values, an SMS alert is sent
to the authorities. In another work [6], the implementation of a water quality monitoring
device combining an Arduino microcontroller and sensors with LabVIEW is presented.
In [7], the water quality is determined with a microcontroller and the Raspberry Pi 3,
coupled to a pH sensor and a TDS sensor. This prediction of water quality is undertaken
with the K-means clustering algorithm based on the values of the measurements. In another
work [8], a water quality measurement system based on an Arduino microcontroller and
sensors using Bluetooth protocol is presented. In papers [9,10], an IoT system based
on ThingSpeak application for water quality monitoring is presented. It consists of an
Arduino Mega and a NodeMCU which are used both for processing the measurements
from the sensors and uploading them to the ThingSpeak server. A wireless sensor network
technology solution for water quality monitoring is described in [11], giving the possibility
of real-time monitoring, where each measurement node has an Arduino microcontroller and
sensors that continuously measure the most important water parameters: pH, temperature,
and conductivity.
On the other hand, the rapid evolution of the Internet of Things (IoT) into a set
of devices for the purpose of management and monitoring requires their connection to
the Internet. For this reason, many networking technologies, especially wireless, have
been used in IoT devices/sensors with specific characteristics and usage, such as Radio
Frequency Identification (RFID), Bluetooth/Bluetooth Low Energy (BLE), Wireless Fidelity
(Wi-Fi), Zigbee, and Long-Range Wide Area Network (LoRaWAN) [12–15]. In many
scientific works, there have been research reports on the Internet of Things (IoT) and
reports examining different ways of implementing it. In [16], the authors provide an
overview of wireless cooperative network communication techniques. A brief overview
of dynamic relaying methods is provided, along with a number of distinct fixed relaying
techniques, such as decode-and-forward. Popular MAC and network level protocols are
the subject of a similar work [17], which also addresses the issue of sensor localization.
In [18], a novel approach to network communication route optimization and IoT node
clustering is discussed. The publications cited above highlight the variety of methods for
optimizing communication while taking into account the limited resources (energy, memory,
and processing power) of Internet of Things devices. The authors of [19] provide an
alternative communication architecture for usage in smart grids and address the importance
of transmission power and packet acknowledgement techniques for extending the lifespan
of battery-powered sensor nodes. Secure authentication in a hierarchical wireless sensor
network is the subject of [20,21]. The performance and security levels of various security
methods are compared by the authors in [22], which provides a complete list of well-known
network security problems that are related to Internet of Things devices. Many research
papers have been published on the transmission of data, including indicative research
works on the LoRa network [23,24], Bluetooth [25–27], wireless network (Wi-Fi) [28–30],
and Zigbee protocol [31–33].
In this paper, an integrated low-cost water quality monitoring system is presented. The
novelty of the proposed system lies in the simplicity of its implementation, the affordability
of the hardware (microcontroller and sensors), and the open-source software used in the
information system to collect and display the measurements in real time. The aim is to
evaluate the behavior of the low-cost sensors as well as to validate the reliability of the
results by investigating a way of correcting the measurements in relation to measurements
Electronics 2025, 14, 857 3 of 17

from reference instruments. For research purposes, the improvement of measurements

for TDS by applying a non-linear equation and for pH by applying a linear equation
was investigated. The system was implemented using a microcontroller including LoRa
networking and water quality sensors such as TDS, temperature, and pH. The proposed
system is a complement to the work [34], which made it possible to measure the water level
in a village water tank on an island in Greece.
The scientific contribution of this work is demonstrated through the system’s eval-
uation. The identification and determination of correction factors (linear and non-linear
equations) were key aspects of the research. Applying these coefficients to correct the
measurements from low-cost sensors resulted in more reliable data, enhancing the value of
the proposed system. Furthermore, implementation with a LoRa network ensures extensive
spatial coverage in an unlicensed frequency band, making it ideal for rural areas. The
results are encouraging, as measurements from low-cost sensors can be made more reliable
by using correction factors in accordance with the proposed methods. By integrating the
present application into the existing system, it is possible to visualize a complete set of
real-time measurements of the drinking water in a reservoir in this area.
This article is organized into the following sections. Section 2 describes the mate-
rials and methods, Section 3 comprises the results and discussion, and Section 4 offers
the conclusions.

2. Materials and Methods

The experiment was carried out in the field and the installation of the device was
undertaken in a water tank in the village of Pirgos on the island of Tinos in Greece (Figure 1).
The tank was made of concrete and its total capacity was 200 cubic meters. This work
is an extension of the LoRa node of [34], since the infrastructure was available from an
earlier work; for this reason, the experiment in the present work was performed on the
same installation. The monitoring station operated from 29 July to 20 August 2024, with
measurement data being taken every ten minutes. For this research, as the data were very
similar, a day’s worth of data were analyzed and tested against calibrated and certified
instruments in sight by hand. The metering system was an autonomous energy system
that used a photovoltaic panel and a battery (Figure 2). The reason for the autonomous
energy is that there was no electricity supply in the area where the water tanks were located.
Given that the microcontroller used met the needs of the extension both in terms of the
interconnection of the additional sensors (analog digital ports, I2C, UART) used and the
computing processing power, it was possible to complete the project with the construction
of an expansion board. The microcontroller used was the TTGO@ESP32 (Shenzhen, China)
(LoRA 868 Mhz) [35] (Figure 2c) and the additional sensors used were a total dissolved
solids (TDS) sensor, a temperature sensor, and a puissance hydrogen (pH) sensor. The
microcontroller was programmed using the Arduino IDE [36], which provides a user-
friendly environment based on a variant of the C programming language. Additionally,
this software offers a wide range of libraries for various sensor types, enabling easy and
rapid project implementation.
Electronics
Electronics 2025,
Electronics2025, 14,
14,x857
2025,14, xFOR
FORPEER
PEERREVIEW
REVIEW 444of
of17
of 17

(a)
(a) (b)
(b)
Figure
Figure 1.1. Village
Village of
Villageof Pirgos
ofPirgos Tinos:
PirgosTinos: (a)(a)
Tinos: map
(a) map
of of
map of the
the village
village
the of
of Pirgos
of Pirgos
village [37];[37];
Pirgos (b) (b)
[37];the the
the water
(b)water tank
tank of
water the of
tank of the
experi-
the
experiment.
ment.
experiment.

(a)
(a) (b)
(b) (c)
(c)
Figure
Figure2.2.
2.Water
Water
Water tank
tank
tank level monitoring
level
level station
monitoring
monitoring [34]:
station
station (a)
(a)the
[34]:
[34]: (a)autonomous
the energy
the autonomous
autonomous water
energy
energy tank
tanklevel
waterwater tankmon-
level level
mon-
itoring
itoringstation;
monitoring (b)
(b)the
station;
station; (b)parts
the of
ofthe
the parts
parts ofmonitoring
the station;
the monitoring
monitoring (c)
(c)microcontroller
station;
station; TTGO@ESP32.
(c) microcontroller
microcontroller TTGO@ESP32.
TTGO@ESP32.

AA totaldissolved
Atotal
total dissolvedsolids
dissolved solidssensor
solids sensor
sensor shows
shows
shows how
howhow many
many many milligrams
milligrams
milligrams of of soluble
ofsoluble
soluble solidssolids
solids are are
aredis-
dis-
dissolved
solved
solvedin in
inone one
oneliter liter
literof of water.
ofwater. Regarding
water.Regarding
RegardingTDS, TDS, if it is
TDS,ififititisishigh, high, it
high,ititmeans means
meansthat that
thatmore more substances
moresubstances
substancesare are
are dissolved
dissolved in in
the the water
water and, and,
in in general,
general, thethe lower
lower the the purity
purity ofof the
the
dissolved in the water and, in general, the lower the purity of the water. Greek law [38] water.
water. Greek
Greek law
law [38]
[38]
sets ◦
setsaaaconductivity
sets conductivitylimit
conductivity limitfor
limit fordrinking
for drinkingwater
drinking waterof
water ofof2500
2500µS/cm
2500 µS/cm
µS/cm atat
at 202020°C,C,
°C, which
which
which is equivalent
isisequivalent
equivalent to
to
to total
total dissolved
dissolved solid
solid values
values of of 1375
1375 toto 1875
1875 mg/lt,
mg/lt, depending
depending on
total dissolved solid values of 1375 to 1875 mg/lt, depending on the water source [39]. on the
the water
water source
source [39].
[39].
Therefore,
Therefore,
Therefore,thisthis measurement
thismeasurement
measurementcan can
canbebe used
beused
usedasas one
asone
oneof of the
ofthe indicators
theindicators
indicatorsto to evaluate
toevaluate
evaluatethe the purity
thepurity
purityof of
of
water.
water. The TDS
TDS sensor
sensor used
used was
was the
theDFROBOT
DFROBOT SEN0244
SEN0244 [40] (Figure
[40]
water. The TDS sensor used was the DFROBOT SEN0244 [40] (Figure 3a), the measure- (Figure3a), the
3a), measurement
the measure-
Electronics 2025, 14, x FOR PEER REVIEW 5 of 17
process
ment
mentprocesswas was
process based
wasbased on water
based on
onwater conductivity,
water conductivity,
conductivity,and and
and thethe sensor
the sensoroutput
sensor output(measurement)
output (measurement)was was
was
provided
provided via
via an
an analog
analog communication
communication
provided via an analog communication interface. interface.
interface.
The
The temperature
temperature ofof water
water affects
affects its
its taste.
taste. As
As the
the temperature
temperature increases,
increases, the
the water
water
becomes less palatable because the gases dissolved in it are expelled. When
becomes less palatable because the gases dissolved in it are expelled. When the tempera- the tempera-
ture
tureofofthe
thewater
waterexceeds
exceeds1515°C,
°C,any
anymicrobes
microbespresent
presentin inthe
thewater
watergrow.
grow.The
Thesensor
sensorused
used
was
was the DFROBOT Waterproof DS18B20 [41] (Figure 3b) with an I2C communicationin-
the DFROBOT Waterproof DS18B20 [41] (Figure 3b) with an I2C communication in-
terface.
terface.
The
The pH pH (puissance
(puissance hydrogen)
hydrogen) indicates
indicates thethe concentration
concentration of of hydrogen
hydrogen ions
ions in
in the
the
water.
water.When
Whenthe thepH
pHvalue
valueisisbelow
below7,7,the
thewater
waterisisacidic;
acidic;when
whenititisis77the
thewater
waterisisneutral.
neutral.
When
When itit isis above
above 77 and
and up
up toto 14,
14, water
water isis classified
classified as
as alkaline.
alkaline. The
The pH pH sensor
sensor used
used was
was
the
theDFROBOT
DFROBOTSEN0161 SEN0161[42]
[42](Figure
(Figure3c);
3c);this
thissensor
sensorisisananelectrochemical
electrochemicalsensor,
sensor,and
andthethe
output (measurement) of the sensor is provided by analog communication
output (measurement) of the sensor is provided by analog communication interface. interface.
(a) (b) (c)

quality: (a) TDS sensor (DFROBOT

Figure 3. The set of sensors used for the measurement of water quality:
SEN0244); (b) temperature
temperature sensor
sensor (Waterproof
(WaterproofDS18B20);
DS18B20);(c)
(c)pH
pHsensor
sensor(DFROBOT
(DFROBOTSEN0161).
SEN0161).

Choosing DFROBOT sensors for various applications offers several advantages, in-
cluding reliability, ease of use, and affordability. These sensors are known for their high
accuracy and seamless integration into IoT systems, as demonstrated by studies analyzing
their use in water quality monitoring and agricultural applications [43,44]. Their quick
Electronics 2025, 14, 857 5 of 17

The temperature of water affects its taste. As the temperature increases, the water
becomes less palatable because the gases dissolved in it are expelled. When the temperature
of the water exceeds 15 ◦ C, any microbes present in the water grow. The sensor used was
the DFROBOT Waterproof DS18B20 [41] (Figure 3b) with an I2C communication interface.
The pH (a) (puissance hydrogen) indicates (b) the concentration of hydrogen (c) ions in the
water. When the pH value is below 7, the water is acidic; when it is 7 the water is neutral.
Figure
When 3. it
Theis set of sensors
above 7 and used up tofor14,thewater
measurement
is classifiedof water quality: (a)
as alkaline. TheTDS pH sensor
sensor(DFROBOT
used was
SEN0244);
the DFROBOT (b) temperature
SEN0161 sensor
[42] (Waterproof
(Figure 3c);DS18B20);
this sensor (c)is
pH ansensor (DFROBOT SEN0161).
electrochemical sensor, and the
output (measurement) of the sensor is provided by analog communication interface.
Choosing
Choosing DFROBOT
DFROBOT sensors
sensors forfor
various
various applications
applications offers
offers several
several advantages,
advantages, in-in-
cluding
cluding reliability,
reliability, ease
ease ofof use,
use, and
and affordability.
affordability.These These sensors
sensors areareknown
known forfortheir
theirhighhigh
accuracy and seamless integration into IoT systems, as demonstrated
accuracy and seamless integration into IoT systems, as demonstrated by studies analyzing by studies analyzing
their
their use
useininwaterwaterquality
qualitymonitoring
monitoringand andagricultural
agriculturalapplications
applications[43,44]. [43,44].Their
Their quick
quick
integration and extensive documentation make it easier
integration and extensive documentation make it easier for users to develop and experi- for users to develop and experi-
ment
ment with
with different
differentapplications.
applications. Since
Sincethethesensors
sensors areare
in in
constant
constant contact
contact with
withwater,
water, their
their
lifespan ranges from six months to one year, requiring periodic
lifespan ranges from six months to one year, requiring periodic replacement during system replacement during sys-
tem maintenance.
maintenance. TheThetotaltotalcostcost of ownership
of ownership (TCO) (TCO)
includesincludes both
both the the initial
initial purchase purchase
cost and
cost and ongoing maintenance
ongoing maintenance expenses. expenses.
Two
Two electromagnetic
electromagnetic valves
valves (Figure
(Figure 4a)4a)
andand a plastic
a plastic container
container (Figure
(Figure 4b) were
4b) were used
used
forfor
thetheneeds of the experiment in terms of water intake from the
needs of the experiment in terms of water intake from the tank and for the measure- tank and for the measure-
ment
ment procedure.
procedure. TheThetwo two electromagnetic
electromagnetic valves
valves were were attached
attached totothetheplastic
plasticcontainer,
container, one one
forforwater filling and the other for water draining. The plastic
water filling and the other for water draining. The plastic container used was from container used was from aa
water
water filter
filterandandwas
was tapered
tapered at at
thethebottom
bottom forfor
thethe complete
complete absorption
absorption of of
thethe
water
water sample.
sample.
Given that the measurements were taken periodically, the
Given that the measurements were taken periodically, the procedure for the measurement procedure for the measurement
waswas asasfollows.
follows. AtAtthethestart
start ofof
thetheprocedure,
procedure, both
both electromagnetic
electromagnetic valves
valves were
wereopened
opened forfor
a short period of time to purge the container with water, then
a short period of time to purge the container with water, then only the first electromagneticonly the first electromagnetic
valve
valve waswas operated
operated toto fillfill
thetheplastic
plasticcontainer
container from
from thethesupply
supply ofof thethetank.
tank.The The operating
operating
time of the first electromagnetic valve was calculated so
time of the first electromagnetic valve was calculated so that the container was filledthat the container was filled withwith
water
water andand thethesensors
sensors werewere inin
thethe
water,
water,thenthenthethemeasurements
measurements were
were taken
taken andandfinally
finally onlyonly
the second electromagnetic valve was opened to drain out the
the second electromagnetic valve was opened to drain out the water. It is important to note water. It is important to note
that
thatanyany residual
residual microplastics
microplastics released
released fromfrom thethe container
container diddidnotnotaffect
affect the
theconductivity
conductivity
ofofthethe
water sample and, consequently, did not impact the
water sample and, consequently, did not impact the measurements. The water measurements. The water filter s
filter’s
(food
(food grade)
grade) plastic container
plastic container waswas placed
placed inina concealed
a concealed place,
place, where
where it would
it would bebeunaffected
unaffected
byby any environmental conditions, especially
any environmental conditions, especially high temperatures. high temperatures.

(a) (b)
Figure
Figure 4. 4. Accessories
Accessories forfor
thethe measuring
measuring procedure:
procedure: (a)(a) electromagnetic
electromagnetic valve;
valve; (b)(b) plastic
plastic container
container
from water filter.
from water filter.

The data via the LoRa network were sent to an information system. In the information
system the data were collected and visualized. For the collection and storage of the
measurements an InfuxDB [45] database was used, while the visualization was undertaken
with the Grafana Lab software [46]. Both the operating system of the information system
Electronics 2025, 14, 857 6 of 17

and the software used are open-source. Grafana is an open-source data visualization
and monitoring platform widely utilized across various domains for its ability to create
dynamic and interactive dashboards. It serves as a powerful tool for visualizing complex
data from multiple sources, including databases such as InfluxDB, Prometheus, MySQL,
and PostgreSQL. Grafana’s versatility allows users to customize panels and dashboards
to suit specific analytical needs, making it a popular choice for real-time data analysis in
diverse applications, from industrial IoT systems to renewable energy monitoring [47–49].
InfluxDB is a powerful time-series database designed specifically for handling high volumes
of time-stamped data, making it particularly suitable for applications in fields such as IoT,
monitoring systems, and data analytics. Its architecture is optimized for the fast ingestion
and retrieval of time-series data, which is crucial for real-time analytics and monitoring
applications. InfluxDB supports an SQL-like query language, which allows users to perform
complex queries on time-series data efficiently [50].

3. Results and Discussion

In this section, the measurements of the water quality through the web interface
of the Grafana software as well as the methodologies used for the normalization of the
measurements are presented. The water quality data are reported on total dissolved
solids (TDS), water temperature, and pH. The evaluation of the results was carried out
by comparing the measurements from the low-cost water quality monitoring system with
the measurements from reference instruments (for TDS the reference instrument is the
Milwaukee TDS Meter MW401 PRO (Milwaukee Electronics Kft, Szeged, Hungary), for
temperature the Hanna Instruments HI 98509 (Hanna Instruments Inc, Woonsocket, RI,
USA), and for pH the Milwaukee pH Meter MW101 PRO, (Milwaukee Electronics Kft,
Hungary). As the monitoring instruments used do not support telemetry, the measurements
were taken in the field (water tank). Finally, the evaluation of the results is presented
through the root mean square error (RMSE), mean average deviation (MAD), and mean
average error (MAE) methods.

3.1. Grafana Interface

The Grafana Lab software was used to visualize the measurements of the proposed
low-cost system, as it offers a user-friendly environment based on a web application. In
Electronics 2025, 14, x FOR PEER REVIEW 7 of 17
addition, the software supports user categorization in terms of user rights. Figure 5 shows
the web interface of the Grafana Lab v9.5.2 software.

Figure 5.
Figure 5. Web
Webinterface
interfaceofof
the Grafana
the LabLab
Grafana software.
software.

3.2. Data Analysis

For the data analysis, the measurements were taken on 12 August 2024 for a period
of eight hours (from 9:00 am to 5:00 pm). Τhe size of the water tank was 200 cubic meters.
Three low-cost sensors were used, namely a TDS, temperature, and pH sensor. Every sam-
ple consisted of three floating point values for each sensor. The first metric was the TDS
Electronics 2025, 14, 857 7 of 17

3.2. Data Analysis

For the data analysis, the measurements were taken on 12 August 2024 for a period
of eight hours (from 9:00 am to 5:00 pm). The size of the water tank was 200 cubic meters.
Three low-cost sensors were used, namely a TDS, temperature, and pH sensor. Every
sample consisted of three floating point values for each sensor. The first metric was the
TDS value, measured in mg/L with an accuracy of a 32 bit float number, the second metric
was the temperature in Celsius with an accuracy of a 16 bit float number, and finally the
pH metric, measured in pH with an accuracy of a 16 bit float with an actual accuracy of 0.1.
The sample size (n) for the evaluation period of the experiment, as described below, was
six samples per hour, with a 10 min interval between each sample, for a total of 8 h. This
adds up to 48 samples (n = 48). As there were not many similar devices operating in this
region, no error rate was observed during transmission and reception. Additionally, no
erroneous measurements, such as outliers or values exceeding limits, were detected. Data
transmission via LoRa from the TTGO@ESP32 was handled using the CayenneLPP library,
where the device address (devAddr), as well as the network (nwkSKey) and application
(appSKey) keys, are defined in the code.
The measured data according to the library (CayenneLPP) specifications must be
entered in specific data channels, therefore the measurement of TDS was in data chan-
nel 37, temperature in data channel 38, and pH in data channel 39. The code has the
following format
payload.reset();
payload.addAnalogInput(37, tds);
payload.addAnalogInput(38, temp);
payload.addAnalogInput(39, ph);
The variables “tds”, “temp”, and “ph” contain the current measurement. Additional
security was implemented using connection credentials before the actual data packet
was transmitted.
For research purposes, the investigation of the correction factor of the TDS mea-
surements of the low-cost sensor was carried out by the linear and non-linear correction
methods presented in works for linear correction [51,52] and non-linear correction [53,54].
Electronics 2025, 14, x FOR PEER REVIEW 8 of 17
The results (data sets) concerning the time series of the raw data of the low-cost system
(TDS-LC) and the reference data for total dissolved solids (TDS) are shown in Figure 6a,
while the scatter plot and the degree of determination (linear and non-linear) of correlation
the scatter plot and the degree of determination (linear and non-linear) of correlation be-
between thetwo
tween the twodata
data sets
sets is shown
is shown in Figure
in Figure 6b,respectively.
6b, 6c, 6c, respectively.

(a) (b) (c)

Figure
Figure 6.6.TDS
TDS measurements
measurements ofof thethe
low-cost system
low-cost and reference
system instrument:
and reference (a) time series
instrument: of TDS
(a) time series of
measurements of the low-cost system and reference instrument; (b) scatter plot
TDS measurements of the low-cost system and reference instrument; (b) scatter plot and the and the degree of degree
determination (linear
of determination equation)
(linear equation)between the measurements
between of TDS of
the measurements of the
TDS low-cost
of the system andsystem
low-cost ref- and
erence instrument;
reference instrument;(c)(c)scatter plot
scatter and
plot thethe
and degree of determination
degree (non-linear
of determination equation)
(non-linear between
equation) between
thethe measurementsofofTDS
measurements TDS ofof the
the low-cost
low-costsystem
systemand
andreference instrument.
reference instrument.

Regarding the raw TDS measurement results shown in Figure 6, the proposed sys-
tem s measurements are higher than the reference values. However, the correlation coef-
ficient is satisfactory (R2 ≈ 0.68), especially considering the use of a low-cost sensor. For
this reason, a linear normalization formula is required, thus ensuring that the measure-
 measurements of the low-cost system and reference instrument; (b) scatter plot and the degree of
determination (linear equation) between the measurements of TDS of the low-cost system and ref-
erence instrument; (c) scatter plot and the degree of determination (non-linear equation) between
Electronics 2025, 14, 857 the measurements of TDS of the low-cost system and reference instrument. 8 of 17

Regarding the raw TDS measurement results shown in Figure 6, the proposed sys-
Regarding
tem s measurementsthe raw TDS
are measurement
higher results shown
than the reference in Figure
values. 6, thethe
However, proposed system’s
correlation coef-
measurements are higher than the reference values. However, the correlation
ficient is satisfactory (R2 ≈ 0.68), especially considering the use of a low-cost sensor. Forcoefficient
is satisfactory
this (R2 ≈ normalization
reason, a linear 0.68), especially considering
formula the use
is required, of aensuring
thus low-costthatsensor. For this
the measure-
reason, a linear normalization formula is required, thus ensuring that the measurements
ments from the low-cost system are more realistic. As a first step in normalizing the val-
from the measurements
ues, the low-cost system are downscaled
were more realistic. As aa first
using step infactor,
correction normalizing
which the
wasvalues, the
multiplied
measurements were downscaled using a correction factor, which was multiplied
by each primary measurement. A series of tests were then conducted, leading to the de- by each
primary measurement.
termination A seriesfactor,
of the correction of tests werewas
which thenfound
conducted, leading
to be 0.8. The to the determination
corrected TDS value
of the correction factor, which
is derived from Equation (1). was found to be 0.8. The corrected TDS value is derived
from Equation (1).
𝑇𝐷𝑆 = 𝑇𝐷𝑆 ∙𝐶 (1)
TDS L−Corrected = TDS Raw · CF (1)
whereTDS 𝑇𝐷𝑆 is the corrected TDS measurement value (by linear equation),
where L−Corrected is the corrected TDS measurement value (by linear equation), TDS Raw
𝑇𝐷𝑆 is the raw TDS
is the raw TDS measurement measurement
value, and value,
CF isand 𝐶 is the correction
the correction factor (CF factor . = 0.8 .
= 0.8)(𝐶
The results
The resultsafter
afterapplying
applying Equation
Equation (1)
(1) to
tothe
theTDS
TDSmeasurements
measurementsfromfromthethelow-cost
low-cost
system are shown in Figure 7. In particular, Figure 7a shows the time series of
system are shown in Figure 7. In particular, Figure 7a shows the time series of the corrected the cor-
rected
TDS TDS measurements
measurements and theand the reference
reference TDS measurements,
TDS measurements, whilewhile Figure
Figure 7b shows
7b shows the
the scatter plot and the degree of correlation between the corrected TDS measurements
scatter plot and the degree of correlation between the corrected TDS measurements and the
and the reference
reference TDS measurements.
TDS measurements.

(a) (b)

Figure 7. Linear corrected TDS measurements of the low-cost system and reference instruments:
(a) time series of linear corrected TDS measurements of the low-cost system and reference instrument;
(b) scatter plot and the degree of determination between the linear corrected TDS measurements of
the low-cost system and reference TDS measurements.

For research purposes, the application of a non-linear equation as a correction factor

was investigated. After repeated tests, a third-degree polynomial equation as a correction
factor showed more realistic results in the measurements. The correction polynomial
appears in Equation (2).

TDSP−Corrected = −0.000001 · TDSRaw 3 + 0.0000351 · TDSRaw 2 + 0.88 · TDSRaw + 1.25 (2)

where TDSP−Corrected is the corrected TDS measurement value (by non-linear equation),
and TDSRaw is the raw TDS measurement value.
Figure 8a shows the time series of the corrected TDS measurements and the reference
TDS measurements by non-linear equation, while Figure 7b shows the scatter plot and
the degree of correlation between the corrected TDS measurements and the reference TDS
measurements by non-linear equation.
 and 𝑇𝐷𝑆 is the raw TDS measurement value.
Figure 8a shows the time series of the corrected TDS measurements and the reference
TDS measurements by non-linear equation, while Figure 7b shows the scatter plot and the
Electronics 2025, 14, 857 degree of correlation between the corrected TDS measurements and the reference9 TDSof 17
measurements by non-linear equation.

(a) (b)

Figure8.8.Non-linear
Figure Non-linearcorrected
corrected
TDSTDS measurements
measurements of the
of the low-cost
low-cost systemsystem and reference
and reference instru-
instruments:
(a) time(a)
ments: series
timeof non-linear
series corrected
of non-linear TDS measurements
corrected of theoflow-cost
TDS measurements system
the low-cost and and
system reference
refer-
instrument; (b) scatter
ence instrument; plot and
(b) scatter plotthe
anddegree of determination
the degree between
of determination the non-linear
between corrected
the non-linear TDS
corrected
measurements of the low-cost system and reference TDS measurements.
TDS measurements of the low-cost system and reference TDS measurements.

According
According to Figures 77and
to Figures and8,8,which
which illustrate
illustrate thethe corrected
corrected TDSTDS measurements,
measurements, alt-
although they show values close to the reference values, it is observed that there are
hough they show values close to the reference values, it is observed that there are residual
residual deviations with respect to the reference measurements. For research purposes
deviations with respect to the reference measurements. For research purposes regarding
regarding the identification of trends indicated by the time series in terms of reliability and
the identification of trends indicated by the time series in terms of reliability and system
system evaluation, the moving average (MA) method (with an interval of 5) was applied
evaluation, the moving average (MA) method (with an interval of 5) was applied to both
to both the corrected (linear and non-linear) measurements of the low-cost system and
the corrected (linear and non-linear) measurements of the low-cost system and the refer-
the reference measurements to investigate the reliability of the results. Figure 9 shows the
ence measurements to investigate the reliability of the results. Figure 9 shows the results
results of applying the moving average to the measurements of both the low-cost (linear
of applying the moving average to the measurements of both the low-cost (linear cor-
corrected) and reference system. Figure 9a shows the time series of the measurements
rected) and reference system. Figure 9a shows the time series of the measurements after
after applying the moving average method to the measurements of both the low-cost
applying the moving average method to the measurements of both the low-cost (linear
(linear corrected) and the reference system, while Figure 9b shows the scatter plot and the
corrected) and the reference system, while Figure 9b shows the scatter plot and the10deter-
Electronics 2025, 14, x FOR PEER REVIEW of to
17
determination of the correlation coefficient after applying the moving average method
mination of the correlation coefficient after applying the moving average method to the
the measurements of both the low-cost (linear corrected) and the reference system.
measurements of both the low-cost (linear corrected) and the reference system.

(a) (b)

Figure9.9.Moving
Figure Movingaverage
averagemethod
method application
application of of low-cost
low-cost system
system measurements
measurements (linear
(linear corrected)
corrected) and
and reference
reference measurements:
measurements: (a) series
(a) time time series of moving
of moving averageaverage TDS linear
TDS linear corrected
corrected measurements
measurements of the
low-cost system and
of the low-cost reference
system instrument;
and reference (b) scatter(b)
instrument; plotscatter
and the degree
plot and of determination
the between the
degree of determination
moving average TDS linear corrected measurements of the low-cost system
between the moving average TDS linear corrected measurements of the low-cost systemand reference instrument.
and refer-
ence instrument.
Figure 10 shows the results of applying the moving average to the measurements of
both Figure
the low-cost (non-linear
10 shows corrected)
the results and reference
of applying the movingsystems. Figure
average to the9a shows the time
measurements of
series of the measurements after applying the moving average method to the
both the low-cost (non-linear corrected) and reference systems. Figure 9a shows measurements
the time
of bothof
series thethe
low-cost (non-linear
measurements corrected)
after applying system and the average
the moving referencemethod
system,towhile
the Figure 9b
measure-
ments of both the low-cost (non-linear corrected) system and the reference system, while
Figure 9b shows the scatter plot and the determination of the correlation coefficient after
applying the moving average method to the measurements of both the low-cost (non-lin-
ear corrected) system and the reference system.
 ence instrument.

Figure 10 shows the results of applying the moving average to the measurements of
Electronics 2025, 14, 857 both the low-cost (non-linear corrected) and reference systems. Figure 9a shows the time
10 of 17
series of the measurements after applying the moving average method to the measure-
ments of both the low-cost (non-linear corrected) system and the reference system, while
shows
Figurethe
9bscatter
showsplot
the and theplot
scatter determination of the correlation
and the determination coefficient
of the after
correlation applyingafter
coefficient the
moving average method to the measurements of both the low-cost (non-linear
applying the moving average method to the measurements of both the low-cost (non-lin-corrected)
system and the system
ear corrected) reference
andsystem.
the reference system.

(a) (b)

Figure10.
Figure 10.Moving
Movingaverage
averagemethod
methodapplication
applicationofoflow-cost
low-costsystem
systemmeasurements
measurements(non-linear
(non-linear cor-
cor-
rected)and
rected) andreference
referencemeasurements:
measurements:(a)
(a)time
timeseries
seriesofofmoving
movingaverage
averageTDSTDSnon-linear
non-linearcorrected
corrected
measurements
measurementsof ofthe
thelow-cost
low-costsystem
systemand
andreference
referenceinstrument;
instrument;(b)
(b)scatter
scatterplot
plotand
andthe
thedegree
degreeofof
determination
determinationbetween
betweenthethemoving
movingaverage
averageTDSTDSnon-linear
non-linearcorrected
correctedmeasurements
measurementsofofthe
thelow-cost
low-cost
system and reference instrument.
system and reference instrument.

The application of the moving average method to both linear and non-linear corrected
The application of the moving average method to both linear and non-linear cor-
data shows much more realistic values in the measurements as well as optimizing the
rected data shows much more realistic values in the measurements as well as optimizing
deviations. It should be noted that although the improvement of the correlation coefficient
the deviations. It should be noted that although the improvement of the correlation coef-
determination is minimal (maximum R2 ≈ 0.69), in the non-linear correction in particular,
ficient determination is minimal (maximum R2 ≈ 0.69), in the non-linear correction in par-
Electronics 2025, 14, x FOR PEER REVIEW
according to the time series of Figure 10a, the corrected measurements are very close11toofthe 17
ticular, according to the time series of Figure 10a, the corrected measurements are very
reference measurements. In this way the improvement of the measurements from low-cost
close to the reference measurements. In this way the improvement of the measurements
sensors is feasible and effective.
from low-cost sensors is feasible and effective.
Figure
Figure 11
11 shows
shows the
the temperature
temperature measurements of the water in the tank. Specifically,
Figure
Figure 11a
11a shows
shows the
the time
time series
series of
of the
thetemperature,
temperature, low-cost
low-cost system
system data,
data, and
and reference
reference
data,
data, while
while the
the scatter plot and degree of correlation determination are shown in Figure 11b.

(a) (b)

Figure 11.
Figure 11. Temperature measurements
measurements ofof the
the low-cost
low-cost system
system and
and reference
reference instrument:
instrument: (a)
(a) time
time
series of water temperature measurements of the low-cost system and reference instrument;
series of water temperature measurements of the low-cost system and reference instrument; (b) (b) scatter
plot andplot
scatter the and
degree
theof determination
degree betweenbetween
of determination the measurements of water temperature
the measurements of the low-cost
of water temperature of the
system and reference instrument.
low-cost system and reference instrument.

According to Figure 11, the results concerning the water temperature measurements
According to Figure 11, the results concerning the water temperature measurements
between the proposed low-cost and reference system are satisfactory, since both time
between the proposed low-cost and reference system are satisfactory, since both time se-
ries show that the values are very close, and the scatter plot shows that the correlation
determination shows a high degree of correlation (R2 ≈ 0.99).
Finally, the measurements related to the pH of the water are shown in Figure 12,
where Figure 12a shows the time series of pH measurements of both the low-cost system
data and the reference data, while Figure 12b,c shows the scatter plots and the degree of
 series of water temperature measurements of the low-cost system and reference instrument; (b)
scatter plot and the degree of determination between the measurements of water temperature of the
low-cost system and reference instrument.
Electronics 2025, 14, 857 11 of 17
According to Figure 11, the results concerning the water temperature measurements
between the proposed low-cost and reference system are satisfactory, since both time se-
series
ries show
show that
that thevalues
the valuesare
arevery
veryclose,
close,and
andthe
thescatter
scatterplot
plotshows
shows that
that the
the correlation
determination shows a high degree of correlation (R22 ≈≈0.99). 0.99).
Finally, the
Finally, themeasurements
measurementsrelated
relatedto to
thethe
pHpH of the
of water are shown
the water in Figure
are shown 12, where
in Figure 12,
Figure 12a shows the time series of pH measurements of both the low-cost system
where Figure 12a shows the time series of pH measurements of both the low-cost system data and
the reference
data data, while
and the reference Figure
data, 12b,c
while shows
Figure the shows
12b,c scatter the
plots and the
scatter degree
plots of correlation
and the degree of
(linear and non-linear, respectively) between the measurements.
correlation (linear and non-linear, respectively) between the measurements.

(a) (b) (c)

Figure 12. pH
pH measurements
measurements of of low-cost
low-cost system
system and
and reference
reference instrument:
instrument: (a)(a) Time
Time series
series of
of water
water
pH measurements
measurements of low-cost
low-cost system
system and
and reference
reference instrument;
instrument; (b)
(b) Scatter
Scatter plot
plot and
and the
the degree
degree of
determination (linear
determination (linear equation)
equation) between
between the
the measurements
measurements of of water
water pH
pH of
of low-cost
low-cost system
system and
reference instrument; (c) Scatter plot and the degree of determination (non-linear equation)
reference instrument; (c) Scatter plot and the degree of determination (non-linear equation) betweenbetween
the measurements of water pH of low-cost system and reference instrument.
the measurements of water pH of low-cost system and reference instrument.

Regarding pH, the measurements from the low-cost system (according to Figure 12)
Regarding pH, the measurements from the low-cost system (according to Figure 12)
are acceptable although they show a small deviation from the reference measurements.
are acceptable although they show a small deviation from the reference measurements.
Since pH is a critical factor for water quality, it is necessary to normalize the measurements
Since pH is a critical factor for water quality, it is necessary to normalize the measurements
in order to make them more reliable. After a series of tests, a measurement correction
in order to make them more reliable. After a series of tests, a measurement correction
equation, including fixed correction factors (CF1 , CF2 ), was developed and applied to each
measurement. The corrected pH measurements are derived from Equation (3).

pHCorrected = ( pHRaw · CF1 ) − CF2 (3)

where pHCorrected is the corrected pH measurement value, pHRaw is the raw pH measure-
ment value, and CF1 and CF2 are the correction factors (CF1 = 1.06, CF2 = 0.5).
After applying Equation (3) to the raw pH measurements from the low-cost system,
the results are shown in Figure 13. Specifically, Figure 13a shows the time series of the
corrected pH measurements and the reference pH measurements, while Figure 13b shows
the scatter plot and the degree of correlation between the corrected pH measurements and
the reference pH measurements.
As is evident, the normalization of the measurements focuses on the fact that the
measurements have to be more realistic, which is evident from Figure 13a in relation
to Figure 12a from the time series of the data. On the other hand, this procedure does
not degrade the quality of the measurements since the coefficient of determination of
Figures 12b and 13b remains constant.
For research purposes, the application of a non-linear correction factor on the raw
pH measurements from the low-cost sensor was investigated. A third-degree polynomial
equation used as a correction factor yields more realistic results in the measurements. The
correction polynomial is shown in Equation (4).
 After applying Equation (3) to the raw pH measurements from the low-cost system,
the results are shown in Figure 13. Specifically, Figure 13a shows the time series of the
corrected pH measurements and the reference pH measurements, while Figure 13b shows
Electronics 2025, 14, 857 the scatter plot and the degree of correlation between the corrected pH measurements and
12 of 17
the reference pH measurements.

(a) (b)

Figure 13. Corrected pH measurements of the low-cost system and reference instruments: (a) time
series of corrected pH measurements of the low-cost system and reference instruments; (b) scatter
plot and
and the
the degree
degreeof
ofdetermination
determinationbetween
betweenthe
thecorrected pH
corrected measurements
pH of of
measurements thethe
low-cost system
low-cost sys-
and reference pH measurements.
tem and reference pH measurements.

AspHis Pevident, the normalization of 3 the measurements 2 focuses on the fact that the
−Corrected = 1.6262 · pHRaw − 33.864 · pHRaw + 235.5 · pHRaw − 540 (4)
measurements have to be more realistic, which is evident from Figure 13a in relation to
Figure pH
where 12aP−from theistime
Corrected series of the
the corrected pHdata. On the other
measurement valuehand, this procedure
(by non-linear does and
equation), not
degrade
TDSRaw is thethe
quality
raw pH of the measurements
measurement since the coefficient of determination of Figures
value.
12b and 13b14a
Figure remains
showsconstant.
the time series of the corrected pH measurements and the reference
TDS For research purposes,
measurements the application
by non-linear equation, of a non-linear
while Figure 14bcorrection
shows the factor onplot
scatter the raw
and
Electronics 2025, 14, x FOR PEER REVIEW 13 of 17
pH measurements from the low-cost sensor was investigated. A third-degree
the degree of correlation between the corrected pH measurements and the reference pH polynomial
equation used as
measurements byanon-linear
correction equation.
factor yields more realistic results in the measurements. The
correction polynomial is shown in Equation (4).

𝑝𝐻 = 1.6262 ∙ 𝑝𝐻 33.864 ∙ 𝑝𝐻 + 235.5 ∙ 𝑝𝐻 540 (4)

where 𝑝𝐻 is the corrected pH measurement value (by non-linear equation),

and 𝑇𝐷𝑆 is the raw pH measurement value.
Figure 14a shows the time series of the corrected pH measurements and the reference
TDS measurements by non-linear equation, while Figure 14b shows the scatter plot and
the degree of correlation between the corrected pH measurements and the reference pH
measurements by non-linear equation.

(a) (b)

Figure 14.14.Non-linear
Figure Non-linear corrected pHmeasurements
corrected pH measurements oflow-cost
of the the low-cost
system system and reference
and reference instru-
instruments:
(a) time
ments: series
(a) time of non-linear
series corrected
of non-linear pH measurements
corrected of theoflow-cost
pH measurements system
the low-cost and reference
system and reference
instruments; (b) scatter plot and the degree of determination between the non-linear corrected pH
instruments; (b) scatter plot and the degree of determination between the non-linear corrected pH
measurements of the low-cost system and reference pH measurements.
measurements of the low-cost system and reference pH measurements.
3.3. RMSE, MAD, and MAE Method Evaluation
3.3. RMSE, MAD,
For the and MAE
purpose Methodthe
of evaluating Evaluation
proposed low-cost system measurements, the Root
Mean
For Squared Errorof
the purpose (RMSE), Meanthe
evaluating Absolute Deviation
proposed (MAD),
low-cost and measurements,
system Mean Absolute Error
the Root
(MAE) methods were applied to both the TDS and pH of ten-minute periodicity
Mean Squared Error (RMSE), Mean Absolute Deviation (MAD), and Mean Absolute Errormeasure-
ments. The evaluation included the investigation of the above-mentioned methods for
(MAE) methods were applied to both the TDS and pH of ten-minute periodicity measure-
both raw and corrected measurements. With regard to the measurements concerning TDS,
ments. The evaluation included the investigation of the above-mentioned methods for
both raw and corrected measurements. With regard to the measurements concerning TDS,
Table 1 shows the RMSE, MAD, and MAE values for the raw, corrected (linear, non-lin-
ear), and moving average corrected (linear, non-linear) measurements.

Table 1. RMSE, MAD, and MAE methods on TDS measurements of low-cost proposed system.
Electronics 2025, 14, 857 13 of 17

Table 1 shows the RMSE, MAD, and MAE values for the raw, corrected (linear, non-linear),
and moving average corrected (linear, non-linear) measurements.

Table 1. RMSE, MAD, and MAE methods on TDS measurements of low-cost proposed system.

RMSE MAD MAE

Raw TDS measurements 10.4 47.5 47.9
Linear corrected TDS measurements 1.3 38 38.3
Non-linear corrected TDS measurements 0.6 29.4 29.4
Moving average of linear corrected TDS measurements 0.7 9.6 10.8
Moving average of non-linear corrected TDS measurements 1.1 6.6 8.4

Observing Table 1, it is evident that the corrected measurements show significant im-
provement. For all three evaluation methods, the RMSE, MAD, and MAE values are lower
for the corrected measurements in both linear and non-linear correction cases compared
with the raw measurements. Additionally, the variation in raw TDS measurements aligns
with the findings in [10]. Furthermore, the evaluation using the moving average method
also yields lower RMSE, MAD, and MAE values for both raw and corrected measurements,
further supporting the validity of the data correction methodology.
Table 2 shows the RMSE, MAD, and MAE values for the raw and corrected pH
measurements.

Table 2. RMSE, MAD, and MAE methods on pH measurements of low-cost proposed system.

RMSE MAD MAE

Raw pH measurements 0.001 0.2 0.18
Linear corrected pH measurements 0.012 0.21 0.19
Non-linear corrected pH measurements 0.0009 0.09 0.10

Regarding pH, Table 2 presents the RMSE, MAD, and MAE values for both raw and
corrected measurements. The raw measurements of pH align with the findings in [8]. The
results of all three evaluation methods for each pH measurement case are satisfactory.
As shown, the differences are very small, which is expected, as the correction does not
significantly enhance the measurements. Instead, its primary purpose is to align the
corrected measurements more closely with the reference measurements.

4. Conclusions
Water quality is nowadays an integral part of both the environment and human health.
Especially for human beings, water is a source of life. The quality of water, especially
potable water, must be checked at regular intervals. Although systems that measure the
quality of drinking water are reliable, in many cases their use is prohibitive as they are
expensive to purchase and costly to maintain. By Greek law [36], the quality of potable
water must be analyzed in a chemical laboratory one or two times per year. In many cases,
particularly in remote rural areas, water is monitored by mobile monitoring units, at long
intervals or not at all. In this work, a real-time water quality monitoring system based on
low-cost devices (accessible sensors and a microcontroller) and data transmission via the
LoRa network was presented.
The innovation of the proposed system relates to both the composition of the low-
cost monitoring station’s construction and the open-source-based information system.
With the implementation of these two factors, the objective achieved is the real-time
monitoring of water quality. Aiming at optimizing the evaluation of the proposed system,
a calibration procedure of the measurements of these sensors was performed. Meanwhile,
Electronics 2025, 14, 857 14 of 17

the LoRa network was a significant milestone in this work as it enabled long-distance
data transmission. Notably, the network demonstrated stability with no packet error rate,
provided extensive spatial coverage, and featured an accessible implementation method,
making it ideal for Internet of Things (IoT) applications.
The proposed methodology shows excellent results, as the results show satisfactory
degrees of correlation determination between the measurements from the low-cost system
and the reference instruments. Specifically, for total dissolved solids (TDS) it showed R2
≈ 0.68, and for pH it showed R2 ≈ 0.69. With respect to temperature, R2 showed a very
high degree of correlation, (R2 ≈ 0.99), and therefore no further investigation was needed.
The application of linear and non-linear correction factors to the raw measurements, while
not resulting in a better coefficient of determination, yielded more realistic measurement
values. The improvement in the measurements was such that the linear equation of the
trend line for the corrected measurements shows lower values for the coefficients (a, b)
in the equation (y = a · x + b) compared with the linear equation of the trend line for the
raw measurements.
On the other hand, the confirmation of the results in terms of corrected measurements
is evident from the Root Mean Squared Error (RMSE), Mean Absolute Deviation (MAD),
and Mean Absolute Error (MAE) methods. These methods were applied to both primary
and corrected measurements for TDS and pH. The RMSE, MAD, and MAE results show
that the both the linear and non-linear corrected measurements show lower or almost
the same values in all the above-mentioned methods, which shows the reliability of the
measurements. In particular, it should be noted that the non-linear correction yielded
better results than the linear correction for both TDS and pH measurements, indicating an
improvement in the reliability of the proposed system.
Obviously, the proposed system cannot replace the official methods and instruments
for measuring drinking water quality; however, its application could be an affordable
and efficient solution, which means it is suitable for small communities in rural areas for
monitoring the quality of potable water in real time.

Author Contributions: Conceptualization, I.C. (Ioannis Chronis) and I.C. (Ioannis Christakis);
methodology, I.G., S.M., S.K., I.C. (Ioannis Chronis) and I.C. (Ioannis Christakis); software, S.M
and S.K.; validation, I.G., S.M., S.K. and I.C. (Ioannis Christakis); formal analysis, I.C. (Ioannis Chro-
nis) and I.C. (Ioannis Christakis); investigation, I.G., S.M. and S.K.; resources, I.G. and S.K.; data
curation, I.C. (Ioannis Chronis) and I.C. (Ioannis Christakis); writing—original draft preparation,
I.G., S.M. and S.K.; writing—review and editing, I.C. (Ioannis Chronis) and I.C. (Ioannis Christakis);
visualization, I.G., S.M. and S.K.; supervision, I.C. (Ioannis Chronis) and I.C. (Ioannis Christakis);
project administration, I.C. (Ioannis Christakis). All authors have read and agreed to the published
version of the manuscript.

Funding: This research received no external funding.

Data Availability Statement: All of the data created in this study are presented in the context of
this article.

Conflicts of Interest: The authors declare no conflicts of interest.

References
1. United Nations, Department of Economic and Social Affairs. Sustainable Development Goals. Available online: https://sdgs.un.
org/goals/goal6 (accessed on 10 December 2024).
2. Lin, L.; Yang, H.; Xu, X. Effects of Water Pollution on Human Health and Disease Heterogeneity: A Review. Front. Environ. Sci.
2022, 10, 880246. [CrossRef]
3. Almaviva, S.; Artuso, F.; Giardina, I.; Lai, A.; Pasquo, A. Fast Detection of Different Water Contaminants by Raman Spectroscopy
and Surface-Enhanced Raman Spectroscopy. Sensors 2022, 22, 8338. [CrossRef] [PubMed]
Electronics 2025, 14, 857 15 of 17

4. Osman, S.O.; Mohamed, M.Z.; Suliman, A.M.; Mohammed, A.A. Design and Implementation of a Low-Cost Real-Time In-Situ
Drinking Water Quality Monitoring System Using Arduino. In Proceedings of the 2018 International Conference on Computer,
Control, Electrical, and Electronics Engineering (ICCCEEE), Khartoum, Sudan, 12–14 August 2018; pp. 1–7.
5. Chowdury, M.S.U.; Emran, T.B.; Ghosh, S.; Pathak, A.; Alam, M.M.; Absar, N.; Andersson, K.; Hossain, M.S. IoT Based Real-Time
River Water Quality Monitoring System. Procedia Comput. Sci. 2019, 155, 161–168. [CrossRef]
6. Taru, Y.K.; Karwankar, A. Water monitoring system using arduino with labview. In Proceedings of the 2017 International
Conference on Computing Methodologies and Communication (ICCMC), Erode, India, 18–19 July 2017; pp. 416–419.
7. Pappu, S.; Vudatha, P.; Niharika, A.V.; Karthick, T.; Sankaranarayanan, S. Intelligent IoT based water quality monitoring system.
Int. J. Appl. Eng. Res. 2017, 12, 5447–5454.
8. Feng, C.; Yuan, J.; Sun, Y.; You, J. Design of Water Quality Monitoring System. In Proceedings of the 2020 International Conference
on Artificial Intelligence and Computer Engineering (ICAICE), Beijing, China, 23–25 October 2020; pp. 264–267.
9. Pasika, S.; Gandla, S.T. Smart Water Quality Monitoring System with Cost-Effective Using IoT. Heliyon 2020, 6, e04096. [CrossRef]
10. Goparaju, S.U.N.; Vaddhiparthy, S.S.S.; Pradeep, C.; Vattem, A.; Gangadharan, D. Design of an IoT System for Machine Learning
Calibrated TDS Measurement in Smart Campus. In Proceedings of the 2021 IEEE 7th World Forum on Internet of Things (WF-IoT),
New Orleans, LA, USA, 14 June–31 July 2021. [CrossRef]
11. Sowmya, C.; Naidu, C.D.; Somineni, R.P.; Reddy, D.R. Implementation of Wireless Sensor Network for Real Time Overhead
Tank Water Quality Monitoring. In Proceedings of the 2017 IEEE 7th International Advance Computing Conference (IACC),
Hyderabad, India, 5–7 January 2017; pp. 546–551.
12. Liu, X.; Yang, Q.; Luo, J.; Ding, B.; Zhang, S. An Energy-Aware Offloading Framework for Edge-Augmented Mobile RFID Systems.
IEEE Internet Things J. 2019, 6, 3994–4004. [CrossRef]
13. Jang, H.; Choe, S.P.; Simon, N.B.G.; Kang, S.; Song, J. A System to Analyze Group Socializing Behaviors in Social Parties. IEEE
Trans. Hum.-Mach. Syst. 2017, 47, 801–813. [CrossRef]
14. Lee, H.-C.; Ke, K.-H. Monitoring of Large-Area IoT Sensors Using a LoRa Wireless Mesh Network System: Design and Evaluation.
IEEE Trans. Instrum. Meas. 2018, 67, 2177–2187. [CrossRef]
15. Varghese, S.G.; Kurian, C.P.; George, V.I.; John, A.; Nayak, V.; Upadhyay, A. Comparative Study of ZigBee Topologies for
IoT-Based Lighting Automation. IET Wirel. Sens. Syst. 2019, 9, 201–207. [CrossRef]
16. Mohammed, A.H.; Dai, B.; Huang, B.; Azhar, M.; Xu, G.; Qin, P.; Yu, S. A Survey and Tutorial of Wireless Relay Network Protocols
Based on Network Coding. J. Netw. Comput. Appl. 2013, 36, 593–610. [CrossRef]
17. Al-Fuqaha, A.; Guizani, M.; Mohammadi, M.; Aledhari, M.; Ayyash, M. Internet of Things: A Survey on Enabling Technologies,
Protocols, and Applications. IEEE Commun. Surv. Tutor. 2015, 17, 2347–2376. [CrossRef]
18. Zhang, C.; Zhang, X.; Li, O.; Yang, Y.; Liu, G. Dynamic Clustering and Compressive Data Gathering Algorithm for Energy-Efficient
Wireless Sensor Networks. Int. J. Distrib. Sens. Netw. 2017, 13, 155014771773890. [CrossRef]
19. Kurt, S.; Yildiz, H.U.; Yigit, M.; Tavli, B.; Gungor, V.C. Packet Size Optimization in Wireless Sensor Networks for Smart Grid
Applications. IEEE Trans. Ind. Electron. 2017, 64, 2392–2401. [CrossRef]
20. Jan, M.; Nanda, P.; Usman, M.; He, X. PAWN: A Payload-Based Mutual Authentication Scheme for Wireless Sensor Networks.
Concurr. Comput. Pract. Exp. 2016, 29, e3986. [CrossRef]
21. Katsoulis, S.; Koulouras, G.; Christakis, I. Energy-Efficient Data Acquisition and Control System using both LoRaWAN and Wi-Fi
Communication for Smart Classrooms. In Proceedings of the 2024 13th International Conference on Modern Circuits and Systems
Technologies (MOCAST), Sofia, Bulgaria, 26–28 June 2024; pp. 1–4. [CrossRef]
22. Mussab, A.; Zaidan, A.A.; Mohammed, T.; Kiah, M.L.M. A review of smart home applications based on Internet of Things. J.
Netw. Comput. Appl. 2017, 97, 48–65.
23. Christakis, I.; Orfanos, V.A.; Chalkiadakis, P.; Rimpas, D. Low-Cost Environmental Monitoring Station to Acquire Health Quality
Factors. Eng. Proc. 2023, 58, 11. [CrossRef]
24. Yuce, M.R. WE-Safe: A Self-Powered Wearable IoT Sensor Network for Safety Applications Based on LoRa: A scoping review. Int.
J. Med. Inform. 2017, 3, 105–108.
25. Raza, S.; Misra, P.; He, Z.; Voigt, T. Building the Internet of Things with Bluetooth Smart. Ad Hoc Netw. 2017, 57, 19–31. [CrossRef]
26. Natgunanathan, I.; Fernando, N.; Loke, S.W.; Weerasuriya, C. Bluetooth Low Energy Mesh: Applications, Considerations and
Current State-of-the-Art. Sensors 2023, 23, 1826. [CrossRef]
27. Zhuang, Y.; Zhang, C.; Huai, J.; Li, Y.; Chen, L.; Chen, R. Bluetooth Localization Technology: Principles, Applications, and Future
Trends. IEEE Internet Things J. 2022, 9, 23506–23524. [CrossRef]
28. Lloret, J.; Sendra, S.; García-Fernández, J.; García, L.; Jimenez, J.M. A WiFi-Based Sensor Network for Flood Irrigation Control in
Agriculture. Electronics 2021, 10, 2454. [CrossRef]
29. Rimpas, D.; Orfanos, V.A.; Chalkiadakis, P.; Christakis, I. Design and Development of a Low-Cost and Compact Real-Time
Monitoring Tool for Battery Life Calculation. Eng. Proc. 2023, 58, 17. [CrossRef]
Electronics 2025, 14, 857 16 of 17

30. Christakis, I.; Moutzouris, K.; Tsakiridis, O.; Stavrakas, I. Barometric Pressure as a correction factor for low-cost particulate matter
sensors. In IOP Conference Series: Earth and Environmental Science; IOP Publishing: Athens, Greece, 2022; Volume 1123, No. 1.
31. Rajesh Gowd, G.; Prasad, M.P.R. ZigBee-Based Health Monitoring System. In Lecture Notes in Electrical Engineering; Springer:
Singapore, 2022; pp. 243–254. [CrossRef]
32. Ercan, S.Ü.; Mohammed, M.S. IoT and XBee Based Central Car Parking Management System. Balk. J. Electr. Comput. Eng. 2022,
11, 35–41. [CrossRef]
33. Haque, K.F.; Abdelgawad, A.; Yelamarthi, K. Comprehensive Performance Analysis of Zigbee Communication: An Experimental
Approach with XBee S2C Module. Sensors 2022, 22, 3245. [CrossRef] [PubMed]
34. Mitropoulos, S.; Orfanos, V.A.; Rimpas, D.; Christakis, I. LoRa Radius Coverage Map on Urban and Rural Areas: Case Study of
Athens’ Northern Suburbs and Tinos Island, Greece. Eng. Proc. 2023, 58, 19. [CrossRef]
35. Santos, S. ESP32 with Built-in SX1276 Lora and SSD1306 OLED Display (Review). Maker Advisor. 2023. Available online:
https://makeradvisor.com/esp32-sx1276-lora-ssd1306-oled/ (accessed on 30 July 2024).
36. Arduino Software. Available online: https://www.arduino.cc/en/software (accessed on 12 February 2025).
37. Google Maps. Available online: https://www.google.com/maps/@37.6395109 (accessed on 18 November 2024).
38. Gazette of Greek Republic 3525/B‘ 25.5.2023, Common Ministerial Decision ∆1(δ)/ΓΠ Mean. 27829/2023, Quality of Water
Intended for Human Consumption in Compliance with the Provisions of Directive (EU) 2020/2184 of the European Parliament
and of the Council of 16 December 2020 (L435/1, 23.12.2020). Available online: https://www.elinyae.gr/ethniki-nomothesia/ya-
d1dgp-oik-278292023-fek-3525b-2552023 (accessed on 10 December 2024).
39. Rusydi, A.F. Correlation between conductivity and total dissolved solid in various type of water: A review. IOP Conf. Ser. Earth
Environ. Sci. 2018, 118, 012019. [CrossRef]
40. What Is Tds Meter—DFROBOT TDS Meter Sensor with Arduino|DFRobot Electronics. Available online: https://www.dfrobot.
com/product-1662.html (accessed on 12 February 2025).
41. Dfrobot.com. Available online: https://www.dfrobot.com/product-689.html?search=DFR0198&page=1&gad_source=1&gclid=
CjwKCAjwodC2BhAHEiwAE67hJMbY2HBzpAboYgDqLxodz84YwV-Yjia5CS17DXd4aiSF6ntrym0DhhoCZ5UQAvD_BwE (ac-
cessed on 30 July 2024).
42. DFRobot ph Sensor for Arduino—Gravity Analog pH Sensor|DFRobot Electronics. Available online: https://www.dfrobot.com/
product-1025.html (accessed on 12 February 2025).
43. Syafirah, M.; Eso, R.; Husein. IoT-Based Vaname Shrimp Pond Water Quality Monitoring Using the Quamonitor Tool. ELECTRON
J. Ilm. Tek. Elektro 2024, 5, 106–116. [CrossRef]
44. Eso, R.; Mokui, H.T.; Arman, A.; Safiuddin, L.; Husein, H. Water Quality Monitoring System Based on the Internet of Things (IoT)
for Vannamei Shrimp Farming. ComTech Comput. Math. Eng. Appl. 2024, 15, 53–63. [CrossRef]
45. Farmer, K.; InfluxData. InfluxData. Available online: https://www.influxdata.com/ (accessed on 12 February 2025).
46. Grafana Labs. Grafana—The Open Platform for Analytics and Monitoring. Grafana Labs. Available online: https://grafana.com/
(accessed on 12 February 2025).
47. Ndukwe, C.; Iqbal, M.T.; Khan, J. An Open Source LoRa Based, Low-Cost IoT Platform for Renewable Energy Generation Unit
Monitoring and Supervisory Control. J. Energy Power Technol. 2021, 4, 1–25. [CrossRef]
48. Ferencz, K.; Domokos, J.; Kovács, L. Cloud Integration of Industrial IoT Systems. Architecture, Security Aspects and Sample
Implementations. Acta Polytech. Hung. 2024, 21, 7–28. [CrossRef]
49. Pulikottil, T.; Estrada-Jimenez, L.A.; Abadía, J.J.P.; Carrera-Rivera, A.; Torayev, A.; Rehman, H.U.; Mo, F.; Nikghadam-Hojjati, S.;
Barata, J. Big Data Life Cycle in Shop-Floor–Trends and Challenges. IEEE Access 2023, 11, 30008–30026. [CrossRef]
50. Nasar, M.; Kausar, M.A. Suitability of Influxdb Database for Iot Applications. Int. J. Innov. Technol. Explor. Eng. 2019, 8, 1850–1857.
[CrossRef]
51. Yano, S.; Koohsari, M.J.; Shibata, A.; Ishii, K.; Mavoa, S.; Oka, K. Assessing Physical Activity and Sedentary Behavior under
Free-Living Conditions: Comparison of Active Style pro HJA-350IT and ActiGraphTM GT3X+. Int. J. Environ. Res. Public Health
2019, 16, 3065. [CrossRef] [PubMed]
52. Faidah, D.Y.; Kuswanto, H.; Sutikno, S. Improving the Accuracy of Rainfall Prediction Using Bias-Corrected NMME Outputs: A
Case Study of Surabaya City, Indonesia. Sci. World J. 2022, 2022, 9779829. [CrossRef] [PubMed]
Electronics 2025, 14, 857 17 of 17

53. Christakis, I.; Tsakiridis, O.; Sarri, E.; Triantis, D.; Stavrakas, I. Nonlinear Regression Approach as a Correction Factor of
Measurements of Low-Cost Electrochemical Air Quality Sensors. Appl. Sci. 2024, 14, 3282. [CrossRef]
54. Patton, A.; Datta, A.; Zamora, M.L.; Buehler, C.; Xiong, F.; Gentner, D.R.; Koehler, K. Non-Linear Probabilistic Calibration of
Low-Cost Environmental Air Pollution Sensor Networks for Neighborhood Level Spatiotemporal Exposure Assessment. J. Expo.
Sci. Environ. Epidemiol. 2022, 32, 908–916. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.