Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Internet of Things 16 (2021) 100429

Contents lists available at ScienceDirect

Internet of Things
journal homepage: www.sciencedirect.com/journal/internet-of-things

An IoT General-Purpose Sensor Board for Enabling Remote


Aquatic Environmental Monitoring
Jarrod Trevathan, PhD a, *, Simon Schmidtke b, Wayne Read, PhD a, Tony Sharp b,
Abdul Sattar, PhD a
a
Institute for Integrated and Intelligent Systems, Griffith University Nathan, Brisbane QLD, 4111
b
Substation 33, 31 Mary Street, Kingston, Logan, QLD, 4114

A R T I C L E I N F O A B S T R A C T

Keywords: The ability to provide near real-time data (e.g., < every 15 minutes) on aquatic environmental
Environmental informatics conditions via remotely deployed sensors is a highly sought-after capability. However, cost and
Arduino-compatible sensors complexity are often significant factors that limit the spatial and temporal coverage of such
Internet of Underwater Things
monitoring systems. Most existing proposals are expensive, complex and/or are un-malleable for
up-cycled e-waste
adaptation to other environmental sensing applications. This paper presents a simple and flexible
open-source IoT (Internet of Things) electronics design for viable near real-time environmental
measurements – specifically tailored to the rigors of aquatic settings. The system provides the
minimal required functionality for reliable remote sensor readings with a focus on low energy
consumption, renewal energy supply, plug and play deployments, and stability over time. The
system development is driven by actual deployment logistics and constraints. We compare three
revisions of the system/circuit and show how we aspired towards the aforementioned goals,
whilst outlining the evolution of the design based on practical experience. A performance eval­
uation of the system is given in terms of functionality, stability, cost and energy consumption. The
IoT platform is at the core of an affordable near real-time aquatic monitoring system that has been
used in multiple water quality studies. The system has been adapted for use in other applications
including water height monitoring and air dust sensing.

1. Introduction

Water covers approximately two thirds of the Earth’s surface. However, only about 3% of this water is considered fresh and less
than 1% is suitable for human use [1, 2]. The management of scarce water resources is critical for ensuring its sustainability. One of the
primary means of water resource management is the ability to monitor various chemical and biological parameters that directly
impacts water quality. The dynamic interplay of these parameters combined with natural and human inputs can dramatically influence

Acknowledgements (If any): Ian Trevathan, Ron Johnstone, Jody Kruger and Tom Stevens.Source of support: Any grants / equipment / drugs,
and/ or other support that facilitated the conduct of research / writing of the manuscript ( including AFMRC project details, if applicable )Australian
Research Council Linkage (LP190101083), Logan City Council EnviroGrants scheme, Seqwater Community Grants and Griffith University Institute
for Integrated and Intelligent Systems.
* Corresponding author:
E-mail addresses: j.trevathan@griffith.edu.au (J. Trevathan), sschm9@hotmail.com (S. Schmidtke), w.read@griffith.edu.au (W. Read), tonys@
substation33.com.au (T. Sharp), a.sattar@griffith.edu.au (A. Sattar).

https://doi.org/10.1016/j.iot.2021.100429
Received 22 February 2021; Received in revised form 21 May 2021; Accepted 26 June 2021
Available online 4 July 2021
2542-6605/© 2021 Elsevier B.V. All rights reserved.
J. Trevathan et al. Internet of Things 16 (2021) 100429

water conditions [3, 4]. However, there are many impediments and logistical challenges for viably monitoring and assessing water
quality [5, 6].
Vast geographical distances, difficulty of terrain and hostility on equipment make conducting any sort of recurrent water quality
monitoring regime challenging. Most traditional approaches require water samples to be collected manually and analysed in a lab­
oratory. Alternately, sensor equipment is placed in situ with logging devices, which are later retrieved for data download [7]. Either
approach is time consuming, logistically expensive, and results in delayed data. As such, there is a need to be able to affordably collect
water quality data in near real-time via remotely deployed sensors – especially in developing countries or for resource-constrained
operations [8].
Remote aquatic environmental data collection (i.e., deployment, management and control of sensors remotely deployed in the field
[9-12]) reduces the amount of time personnel must spend in the field, thereby limiting the physical dangers and logistical costs [13].
Furthermore, the capacity to view the data in near real-time (i.e., every 5 - 15 minutes) allows decision makers to monitor an event as it
unfolds (e.g., a harmful algal bloom, coral bleaching) and take appropriate counter measures. While such technology now exists to
monitor aquatic/marine environments using remotely deployed networked sensors, expense is still the most significant factor that
limits spatial and temporal coverage. Some examples of aquatic/marine monitoring initiatives include [14-22]. However, most of these
initiatives have since concluded due to the excessive outlay or other reasons. Furthermore, these systems typically have complex
designs with substantial energy requirements. They are also difficult to configure, deploy and maintain without a team of technical
specialists.
The Cave Pearl Project [23] is a proposal for affordable aquatic environmental monitoring. This project uses off-the-shelf com­
ponents (microcontrollers and sensors) to construct inexpensive underwater data loggers for caves. The system runs on three AA
batteries and can log data for approximately one year. However, this system does not provide telemetry for remotely deployed sensor
readings and therefore does not facilitate any of the aforementioned logistical benefits or timely interpretation of the data. Further­
more, the logger’s power source is finite and non-renewable. Additionally, the system is designed for electronics enthusiasts and is not
plug and play from the perspective of ease-of-use for someone with a non-technical background.
This paper presents a simple and flexible open-source IoT (Internet of Things) platform for affordable remote environmental sensing
that extends upon and supersedes the premise of the Cave Pearl Project [23]. The design aspires towards minimal system complexity,
low energy consumption, renewable power supply, plug and play operation and stability/reliability over time using commercial-grade
sensors. The application is informed by practical requirements based on actual deployment logistics and constraints specific to aquatic
environments. Three revisions of the platform are presented showing how the design evolved based on practical experience. A per­
formance evaluation of the system is given in terms of functionality, stability, cost and energy consumption. The IoT platform has been
developed in conjunction with a social enterprise [24-26] and used in multiple environmental studies involving numerous types of
water bodies [27, 28]. The design of the circuit is flexible and adaptable allowing it to be used in other environmental monitoring
applications such as flood level observations and air dust sensing.
This paper is structured as follows: Section 2 provides background and related work. Section 3 presents three designs for IoT aquatic
environmental monitoring showing how each iteration builds on the previous based on evolving requirements. Section 4 provides a
performance evaluation of the designs; and Section 5 provides concluding remarks and avenues for future work.

2. Background, Related Work and Design Aims

This section describes previous monitoring initiatives, water quality parameters, and examines the Cave Pearl Project (which is a
significant influence on the early design of our IoT platform).

2.1. Previous Water Quality Monitoring Initiatives

There have been multiple large-scale and big budget projects for undertaking aquatic and marine environmental monitoring
[14-22]. However, most of these projects are discontinued due to exhausting all of their funding sources and/or political support.
Furthermore, these systems were typically complicated, power intensive, did not have reliable telemetry methods or were unsta­
ble/unreliable. Additionally, most initiatives required a significant team of specialists to configure, deploy and maintain the system.
This often led to a natural conclusion as technology dated, budgets were expended, and physical systems degraded over time.
This paper’s focus is on practical, simple and viable approaches to water quality monitoring through telemetry-based sensing
suitable for developing countries, researchers and small/medium operations [8]. Since 2010, many initiatives have arisen that seek to
use a non-proprietary method to construct viable aquatic environmental monitoring sensor networks (see [27-39]).
Sadler et al. (2015) [29] proposed an environmental monitoring system that interfaced an Arduino with a DHT22 Humid­
ity/Temperature sensor and SDI-12 pressure transducer. The system was powered by a 4,000 mAh lithium ion battery and 6 W solar
panel, and data was transmitted via a General Packet Radio Services (GPRS) modem to a server. Sadler et al. [29] also called for an
open-source, open-data approach to development. However, a physical buoy was not constructed for their system and no actual field
validation tests were conducted in water.
Velez et al. (2015) [30] discuss deploying a swarm of surface drones for border patrol and environmental monitoring. They propose
a low-cost approach using XBee communication modules. Their proposal lacks a solid proof-of-concept, deployment, testing and
commercial considerations.
Lockridge et al. (2016) [31] presented a low-cost, drifting sonde that remotely transmits environmental sensor and geolocation
parameters. The sonde used an Arduino microcontroller and took measurements on temperature and conductivity over several weeks

2
J. Trevathan et al. Internet of Things 16 (2021) 100429

while adrift in the ocean. Arduino’s simplicity and open source nature makes it attractive for environmental monitoring applications.
The system did not use solar power and therefore had limited deployment time.
Perez et al. (2017) [32] designed a low-cost device for monitoring near-shore, shallow marine environments. Their solar-powered
sonde recorded temperature, water pressure and atmospheric pressure remotely using a light-weight, marine-robust platform. They
deployed the system over several months in the Mediterranean Sea. The authors did not outline the exact telemetry method.
Gunawardena et al. (2018) [33] discuss low-cost approaches for environmental monitoring via distributed sensor networks. They
address how open-source microcontrollers (such as Arduino) can be interfaced with commercial-grade sensors. Their proposal eval­
uates energy-constrained applications for weather/atmospheric monitoring but does not address aquatic environments.
Demetillo, Japitana, & Taboada (2019) [34] present a low-cost solution for real-time aquatic environmental monitoring for
shallow, near-shore and calm conditions. The system uses a microcontroller and electrochemical sensors (pH, dissolved oxygen and
temperature). A limited trial deployment of the system is discussed. However, the system has not been tested on a large scale.
Wickert et al. (2019) [35] developed an open-source data logger based on an Arduino microcontroller. The authors state that these
devices are low-cost, lightweight, low-power, and can store up to 32GB of data. Their firmware library is compatible with standard
environmental sensors. Their logger has measured parameters linked to weather, stream flow, and glacier melt across various field
sites. However, these devices are limited to data logging only (i.e., no telemetry).
Horsburgh et al. (2019) [36] describe a web-based, open source data logging system to enable citizen scientists to stream sensor
data from Arduino-based data loggers. Sensor nodes (connected via cellular or Wi-Fi) can post data to a web portal via a web service,
which is stored using Observations Data Model1. This system supports wireless water quality monitoring stations (developed by the
EnviroDIY community) deployed in the Delaware River Watershed.
Table 5
Hadi et al. (2019) [37] propose a device that can remotely monitor the solar panel voltage, current, power, temperature, humidity
and internal battery voltage in real-time. The authors claim battery power savings between 65 – 69%. However, no actual field trials of
this system are reported.
Abdelal, Alsmadi & Jaber (2019) [38] present low-cost solutions for environmental and water resource monitoring. These ap­
plications monitored water evaporation, solar panel power and floodwater. The authors claim that system cost is 5%–10% of com­
parable commercial systems.
Domínguez-Brito et al. (2020) [39] highlight a do-it-yourself approach for low-cost wireless wind data acquisition. The system was
deployed to analyse the foredune of Maspalomas (Canary Islands, Spain). However, there has not been any follow-up application of
this work.
Xi et al. (2019) [40] propose an approach for self-powered environmental sensing buoys that use water wave energy. Zhang, Cao &
Wang, Y. (2019) [41] discuss how low-power wide area networks based on long-range 2.4 GHz communications may be used for
aquatic environmental monitoring. Przybysz et al. (2020) [42] propose a remote deployed sensing buoy designed to work in maritime
environments that communicates via a cellular network. Guan et al. (2021) [43] presents an architecture for a grid of maritime buoys
that collectively form a large cellular network.

2.2. Water Quality Parameters

Water quality is a complex topic and the notion of what constitutes “quality” depends on the aspect on which water quality is being
measured. For simplicity, we will classify water quality according to the following applications: 1) Drinking water; 2) Natural systems;
and 3) Wastewater.
What might constitute acceptable “quality” parameters for wastewater would be highly detrimental if those same conditions
existed in natural systems. Likewise, suitable water quality in a natural system may not meet the minimum requirements set by water
authorities for safe drinking water. This paper focusses on monitoring water quality in natural waterways.
There are seven main indicators of water quality that can be measured by electronic sensors:

1 Temperature – The thermal properties of the water (degrees Celsius/ Fahrenheit);


2 Potential of Hydrogen (pH) – How acidic or basic the water is (between 1 (acidic), 7 (neutral) or 14 (basic));
3 Dissolved Oxygen (DO) – How much oxygen is present in the water (milligrams per litre (mg/L));
4 Conductivity (Salinity) – The ability of the water to pass and electrical current, indicating the water’s salt content (Siemens per metre
(S/m) or micro Siemens (μS/cm) per centimetre);
5 Light – The amount of light that penetrates the water to influence photosynthesis (lux, infrared, lumens, etc.);
6 Turbidity – How much suspended sediment is contained in the water (Nephelometric Turbidity Units (NTU)); and
7 Nutrients:

a. Nitrogen – Animal waste, fertilizer run-off, industrial discharge (milligrams per litre (mg/L)). b. Phosphorus – Similar sources as
nitrogen in addition to decomposing plant matter and aging waterways (milligrams per litre (mg/L)).
Detailed descriptions of these indicators, the sensor technologies, and the hydrodynamic aspects of how they interact are outside of

1
www.odm2.org

3
J. Trevathan et al. Internet of Things 16 (2021) 100429

the scope of this paper. Various international standards govern the acceptable parameters which these indicators should fall within to
support plant and fish life in natural aquatic ecosystems [44-46].

2.3. The Cave Pearl Project

The Cave Pearl Project [23] uses Arduino microcontrollers to construct inexpensive and low-power consumption logging devices to
monitor cave water conditions. Arduino is chosen over other platforms (such as Raspberry Pi) due to its simplicity and low-power
utilisation. The loggers use Arduino-compatible sensors, custom-made logger housings and the minimum components necessary to
construct a logger to keep the design modest and inexpensive (see Table 1).
Figure 1 A illustrates the wiring for the data logger system. The Arduino Pro Mini is connected to the DS3231 RTC (Real-Time Clock)
via the I2C (Inter-Integrated Circuit) protocol. In between logging, the Arduino Pro Mini enters a sleep cycle. An alarm is set on the RTC
to wake the system up every 15 minutes via an electrical signal to pin 2 on the Arduino Pro Mini. An SD (SanDisk) Card Reader/Writer
is connected to the Arduino Pro Mini via the SPI (Serial Peripheral Interface) bus. When the system is awake it reads from the sensors
(connected on the I2C bus) and writes this data to the SD card. Various resistors are added to stabilise the system and built-in LEDs (light
emitting diodes) are removed to conserve power. Wires are soldered directly to the pins on each component to reduce cost. The system is
mounted in a PVC (polyvinyl chloride) housing (Figure 1 B). The data logger costs approximately $100 USD (excluding sensors) and can
run for approximately one year.
While the Cave Pearl Project is noble in its aspirations (i.e., low-cost, simplicity and low-power consumption), the system is limited
to data logging only. The logger must be retrieved from the site to download the data. Furthermore, the power source is non-renewable
which limits the amount of time the logger can be deployed. Finally, the system is designed for electronics enthusiasts and is not plug
and play from the perspective of someone with a non-technical background.

2.4. Design Goals for Viable Near Real–Time Aquatic Monitoring

We propose the design goals and criteria for developing viable remotely-deployed near real-time aquatic monitoring platforms to
be as follows:

1 Simple design – The system architecture should only have the minimal required functionality in order to achieve the required goal
of environmental monitoring;
2 Low-power consumption – Energy is limited once deployed, so the system must be efficient;
3 Renewable power supply – The battery voltage can be recharged by solar energy quicker than it is drained by the system operation;
4 Plug and Play – Deployment and operated/maintained require limited technical expertise;
5 Near real-time – The platform must transmit sensor readings approximately every 15-minutes (or sooner) back to the end user via
the Internet (see [27, 28]);
6 Reliable and accurate sensor data – Sensor data should be reliable and accurate enough to be indicative of change without
necessarily aspiring towards excessive precision;
7 Stable/Reliable – The system should ensure sustained correct operation and gracefully recover from errors/system failures when
deployed over a prolonged period;
8 Affordable – The system needs to be affordable so that it can facilitate viable wide-scale monitoring. (A common misconception is
that low-cost means no-cost.); and
9 Flexible – The platform can be adapted for other environmental monitoring applications.

Figure 2 A illustrates the basic concept for the IoT aquatic environmental monitoring platform. The system is targeted for use in
water bodies that are calm and relatively close to urban centres (as opposed to hostile offshore oceanic conditions). Independent buoys
(i.e., monitoring units) made from existing off-the-shelf components are placed in a water body (Design Goal 1). Each buoy has its own
efficient power supply and solar panels (Design Goals 2 and 3). Deployment can be undertaken by non-technical personnel and does not
require on-site configuration (Design Goal 4).
The platform uses the existing network infrastructure as maintained by the telecommunications companies (i.e., GSM (Global
System for Mobile communication)) to transmit sensor readings via the Internet to an end user (Figure 2 B) (Design Goal 5). ThingsBoard2
is an open source IoT dashboard that is used as the front-end user interface. The sensors are integrated from commercially available
sources and the accuracy, precision and reliability are based on the vendors’ reputations (Design Goal 6). The system performs the basic
operations of sensor sampling, transmission and duty cycling repeatedly and reliably over an extended period of time independently in
the field (Design Goal 7).

3. Developing a Platform for Remote Aquatic Environmental Monitoring

This section outlines how we took the initial hardware concepts from the Cave Pearl Project and modified/enhanced them towards

2
thingsboard.io

4
J. Trevathan et al. Internet of Things 16 (2021) 100429

Table 1
Cave Pearl Data Logger Electronic Components.
Quantity Component Manufacturer

1 3.3 V Arduino Pro Mini 328 Arduino


1 DS3231 RTC Maxim Integrated
1 SD card logger SanDisk
1 CR2032 button cell battery
1 Status LED
1 Battery pack
3 AA alkaline batteries
3 Sensors (light, temperature and pressure) Adafruit
Miscellaneous electronics components and connectors Dupont and Deans

Figure 1. The Cave Pearl Data Logger system (pictures adapted from [23]).

Figure 2. The IoT aquatic environmental monitoring system.

the goals of creating a remote IoT aquatic monitoring system that is affordable, stable, low-power, adaptable and can provide
reasonably accurate sensor data in near real-time. Three revisions of the system are presented. Note that the software, back-end web
data management and user interface are not addressed in this paper (this is the subject of future work).

3.1. First Attempt at Designing a Remote Aquatic Environtmental Monitoring Platform

Table 2 presents the electronic components used in the first version of the aquatic environmental monitoring platform. This attempt
sought to extend the Cave Pearl Data Logger’s components to provide solar charging (solar panels and CN3083 SOP8 charging circuit)
and remote telemetry (Fona 2G GSM module). The platform provides lux (TSL2561) readings at the surface and 0.5 metres underwater,
temperature (MCP9808) and turbidity (Gravity Analog Turbidity Sensor) readings 0.5 metres underwater. (Note that the SD card
logger is not used.) The “Different” column indicates the components that were different to the Cave Pearl Project.
The choice of the three sensor parameters (i.e., lux, temperature and turbidity) was to provide indicative rudimentary water quality
readings without becoming overly focused on the sensor technologies and water quality study application at this stage. Instead, the
goal was to develop a robust system capable of functioning remotely for extended periods of time. The two light sensors (surface and

5
J. Trevathan et al. Internet of Things 16 (2021) 100429

Table 2
First version remote environmental monitoring platform electronic components.
Quantity Component Manufacturer Different

1 5 V Arduino Pro Mini 328 Arduino


1 DS3231 RTC Maxim Integrated
1 CR2032 button cell battery Panasonic
1 Fona 808 Mini Cellular GSM Adafruit
1 2 dBi Mini GSM quad band antenna Adafruit
1 DC 3-32 V step up to 5-35 V boost convertor voltage regulator
8 Lithium ion battery Recycled
4 6 V 0.8 W solar panel
1 CN3083 SOP8 high efficiency solar energy charging circuit chip Texas Instruments
1 MCP9808 Temperature sensor Adafruit
2 TSL2561 Lux sensor Adafruit
1 Gravity Analog Turbidity Sensor DF RobotTM
1 Status LED
1 Miscellaneous electronics components and connectors Dupont and Deans

underwater) provide a means to calculate a light attenuation curve (i.e., surface light penetration through water). Underwater tem­
perature allows us to examine the thermal interactions of surface water mixing. Turbidity provides an indication of suspended sedi­
ment in the water. This sensor required extensive testing and calibration (discussed further in Section 3.3).
Light penetration is an important water quality indicator, as plant and fish species need a certain amount of light. Most commonly
this is measured as the intensity of light in the photosynthetic active radiation (PAR) range (400 – 700 nm) of the electromagnetic
spectrum. Low light penetration could signal the presence of surface weeds, harmful algae blooms, or high amounts of suspended
particulate matter. Temperature is a critical factor that determines the survivability of life in water bodies. Water that is too cold or too
warm can negatively influence an aquatic ecosystem. Turbidity is related to the light penetration and gives an estimate of the total
suspended sediment in a water body. High levels of turbidity may signal excessive nutrient run-off from land.
We selected sensors for these parameters that were inexpensive, provided indicative readings and were compatible with the system
(i.e., interfaced with Arduino via I2C or serial) (Figure 3)s. At this stage, other parameters such as pH, DO, conductivity and ORP
(oxidisation reduction potential) require additional hardware to interface the sensor with our circuitry. Compatible sensors do exist (see
Atlas Scientific3), but their integration with this system is outside of the scope of this paper. We will outline their inclusion in future
work. In this paper, the focus is on providing sufficient sensors to develop the core sensing platform – as an extension of the work from
the Cave Pearl Project [23] and in accordance with our project design goals as outlined in Section 2.3 [27, 28].
Figure 4 A demonstrates the various states of the buoy based on the timing. Upon initially turning on the buoy by connecting it to
the power system, the buoy will cycle between two major states – Awake or Sleep. When an Awake state is triggered, the microcontroller
will enter into a Startup state and initialise all system objects in memory and check all attached hardware. Next the system will take
sensor samples in the Sample state, then enter in to the Transmit state to send the readings to the web server. If an error occurs, the
system will momentarily enter the Error state and notify the user either via a red status LED indicator, transmission of status infor­
mation, or writing an error to the Arduino’s Electrically Erasable Programmable Read-Only Memory (EEPROM). Finally, the system will
enter a Set Alarm state to set the alarm on the RTC. The system then enters into the Sleep state until the alarm wakes the system up again
and the process repeats.
The system takes an “all-or-nothing” approach to transmission. That is, either the data is sent, or it is lost. The versions of the buoy
outlined in this paper do not buffer the data beyond the current duty cycle. If after three unsuccessful attempts at transmission, the
system goes back to sleep and discards the data. Tests and deployments indicated that most failures are due to the system not
establishing a GSM connection rather than failing to transmit. This happens on approximately 1/96 duty cycles a day (a 1.04% error
rate). The conclusions section outlines how we sought to resolve this transmission/buffering issue in a future revision of the buoy.
Figure 4 B illustrates the wiring for the remote environmental monitoring platform. Similar to the Cave Pearl Data Logger, the RTC
and sensors are connected via the I2C bus. The RTC wakes up the system every 15 minutes. Once the sensors have taken their readings,
the Adafruit Fona 2G GSM module transmits these readings to a cloud-based server via the GSM network. 8 recycled lithium ion
batteries power the system (providing 20 Ah). These batteries are charged via an array of four separate 6 V 0.8 W solar panels
(connected in parallel). A solar charging circuit (CN3083) was added to interface the solar panels with the battery. A DC-DC convertor
was included to step the voltage from batteries up to a stable 5 V for the Arduino.
A 3D printed mounting was developed to contain the electronics components (Figure 4 C). Dupont connectors were used on all of
the wires to bridge the connections of the electronics components. The electronics were housed in a PVC canister connected via a flange
to a shaft down to the underwater sensor head (Figure 4 D). Buoyancy was added to the PVC canister by an external foam float (which
the solar panels were mounted on). A 3D printed lid was constructed with a dedicated O-ring housing and IP68-rated cable connector
for the solar panel wiring.

3
atlas-scientific.com

6
J. Trevathan et al. Internet of Things 16 (2021) 100429

Figure 3. Sensors used in the IoT aquatic environmental monitoring platform.

Figure 4. First version electronics schematic, internal housing and deployment.

Five aquatic environmental monitoring buoys were constructed. These were deployed in Lake Ellerslie throughout July/August
2017 (Figure 3 E). The deployment lasted until the 2G GSM signal was completely turned off in Australia. Table 3 shows the outcomes
for this deployment in terms of the duration and whether any of the buoys became challenged or failed.

3.2. Second Attempt at Designing a Remote Aquatic Environtmental Monitoring Platform

The second buoy version addressed the previous version’s shortcomings. These issues include:

7
J. Trevathan et al. Internet of Things 16 (2021) 100429

Table 3
First version environmental monitoring platform deployment outcomes.
Buoy Id Location Deployment Duration Failed / Challenged Reason

Buoy 1 Lake Ellerslie 35 days


Buoy 2 Lake Ellerslie 35 days
Buoy 3 Lake Ellerslie 35 days
Buoy 4 Lake Ellerslie 4 days Leakage – incorrectly fitted O-ring
Buoy 5 Lake Ellerslie 23 days Battery drain

1 The electronics mounting is cumbersome with excessive wiring;


2 There was an ongoing quiescent power draw from the electronics components (i.e., DC-DC convertor) that could not be put to sleep.
This gradual draining of the batteries exceeded their ability to charge via the solar panels;
3 The 2G GSM network was no longer supported; and
4 The physical design of the buoy has floatation, logistic and aesthetic shortcomings.

Table 4 presents the electronic components used in the second version of the buoy. The most significant change to this version was
incorporating all of the components into a concise PCB (Printed Circuit Board). This removed all of the excessive wire connections
between components. Phoenix connectors were used to connect the sensor wires to the PCB. The “Different” column indicates what
components changed between this version and the previous version.
The next issue to be resolved was the power management circuitry. Firstly, the CN3083 solar charging circuit was integrated into
the PCB. A solar charge status LED was added to allow for easy visual inspection of whether the circuit was charging. Next, we
developed an approach to completely power the system down between duty cycles. A JK flip-flop was used for power alternation based
on RTC alarm timing. When the RTC alarm is triggered, a pulse is sent out on its SQW pin. This forces the flip-flop to change states and
power up the system (via a MOSFET (metal-oxide-semiconductor field-effect transistor)). Upon execution of the buoy code a signal is sent
to the JK flip-flop again causing it to change states cutting power to the system.
As 2G was no longer supported in Australia, this version of the buoy needed to be updated to a 3G GSM shield. At the time, Adafruit
had an experimental 3G Fona GSM module on the market. However, this was very expensive and there was limited support for 3G data
in the Adafruit 3G Fona library. Instead, we changed to a TinySine 3G GSM shield, which is a less expensive variant of the Adafruit 3G
Fona GSM shield. We were then able to update the 3G library to provide adequate support for sending HTTP data via 3G.
Figure 5 A illustrates the major electronics components for the second version. This version used a marine polystyrene float for
buoyancy (Figure 5 B). Cutting the polystyrene warranted the development of a hot-wire cutting jig for precise cuts (i.e., hollowing the
centre and cutting a chord across the top). A new 3D printed lid was developed along with an O-ring flange, which held the lid in place
by six M4 hex screws. A sensor head PCB and 3D printed sensor head encapsulating the lux, temperature and turbidity sensor was
developed.
Ten buoys were constructed with five of these deployed for over three months in Lake Ellerslie during 2018 (Figure 5 C). Table 4
shows the outcomes for this deployment in terms of the duration and whether any of the buoys became challenged or failed. The
platform formed the basis for an environmental informatics challenge involving 70 high school students from ten schools (refer to
[28]). The challenge was undertaken as part of a Logan City Council EnviroGrant4 to engage the community in interpretation and
visualisation of aquatic environmental data.

3.3. Third Attempt at Designing a Remote Aquatic Environtmental Monitoring Platform

The third version’s goal was to address the shortcomings of version two, including:

1 Under certain circumstances electrical interference between the RTC and I2C timing can prevent the Arduino from detecting the
wakeup signal (resulting in an infinite sleep cycle);
2 At low battery voltages, the DC-DC convertor becomes unstable resulting in power spikes of up to 35 V being sent through the
system – which destroys vulnerable components;
3 There was humidity build-up inside the buoy over time;
4 There were physical issues with accessing the buoy due to removing multiple screws;
5 Having the electronics free sitting (not secured) in the buoy presents practical problems;
6 The Arduino Pro Mini is limited to 2 KB of Static Random-Access Memory (SRAM); and
7 The turbidity sensor was not integrated correctly or calibrated.

Table 6 presents the electronic components for the third version. The circuit board on this version is referred to as the General-
Purpose Sensor Board (GPSB) to reflect that the system had evolved to a point where the circuit could be used in numerous environ­
mental monitoring applications – not just the aquatic domain. The “Different” column indicates what components changed between

4
logan.qld.gov.au

8
J. Trevathan et al. Internet of Things 16 (2021) 100429

Table 4
Second version remote environmental monitoring platform electronic components.
Quantity Component Manufacturer Different

1 5 V Arduino Pro Mini 328 Arduino


1 DS3231 RTC Maxim Integrated
1 3G GSM Shield TinySine
1 2 dBi Mini GSM quad band antenna Adafruit
1 CR2032 button cell battery
1 DC 3-32 V step up to 5-35 V boost convertor voltage regulator
4 Lithium ion battery Recycled
1 6 V 1 W Solar Panel
1 CN3083 SOP8 high efficiency solar energy charging circuit chip Texas Instruments
1 MCP9008 Temperature sensor Adafruit
2 Adafruit TSL2561 Lux sensor Adafruit
1 Gravity Analog Turbidity Sensor DF RobotTM
1 Power switch
1 Tricolour RGB status LED
Miscellaneous electronics components and connectors Phoenix

Table 5
Second version environmental monitoring platform deployment outcomes.
Buoy Id Location Deployment Duration Failed / Challenged Reason

Buoy 1 Lake Ellerslie 32 days Sensor head water ingress


Buoy 2 Lake Ellerslie 21 days Sensor head water ingress
Buoy 3 Lake Ellerslie 48 days Sensor head water ingress
Buoy 4 Lake Ellerslie 42 days Sensor head water ingress
Buoy 5 Lake Ellerslie 35 days Sensor head water ingress

Figure 5. Second version internal electronics and deployment.

this version and the previous version.


To address the RTC/ I2C conflict that could result in infinite sleep, an overhaul of the buoy’s sleep cycle was required. The RTC was
replaced by a hardware solution using a 555 timer and a 12-bit binary counter set up to cycle the most significant bit every 15 minutes.
Figure 6 A the new sleep and failsafe timing approach for the GPSB. The most significant bit triggers the wake-up signal that wakes the
buoy up. When the buoy finishes executing its code, it sends a sleep command to re-initiate the timer. If this signal is missed, a software
watchdog timer will attempt to send the signal again. If the signal is again missed, the buoy will stay awake for at most 15 minutes

9
J. Trevathan et al. Internet of Things 16 (2021) 100429

Table 6
Third version remote environmental monitoring platform electronic components.
Quantity Component Manufacturer Different

1 5V Arduino Mega 2560 Arduino


1 TinySine 3G GSM Module TinySine
1 2 dBi Mini GSM quad band antenna Adafruit
1 DC 3-32 V step up to 5-35 V boost convertor voltage regulator
6 Lithium ion battery Recycled
2 6 V 1 W solar panel
1 CN3083 SOP8 high efficiency solar energy charging circuit chip Texas Instruments
1 MCP9808 Temperature sensor Adafruit
2 TSL2561 Lux sensor Adafruit
1 Turbidity sensor Custom-built
1 DHT22 Temperature/Humidity sensor Aosong Electronics
1 Power switch
1 Tricolour RGB status LED
Miscellaneous electronics components and connectors Phoenix

Figure 6. General-Purpose Sensor Board internal electronics and deployment.

before a secondary hardware timer kicks in and forces the buoy into sleep. (Note that timestamping of the data now occurs when the
server receives it and the data is entered into the database.)
A DHT22 Humidity / Temperature sensor is included to provide information about internal humidity build-up and to detect leaks.
Desiccant was added to help manage humidity. The buoy lid was also back filled with epoxy resin to reduce the potential for any water
ingress (Figure 6 B). Additional status information regarding the buoy’s operation (uptime, transmit time, network signal strength,
solar charge and battery voltage) was transmitted with the sensor readings.
One of the most notable changes in this version was to use an Arduino MEGA 2560 (Figure 6 C). The MEGA 2560 has the same basic
processor as the Pro Mini, but the SRAM is doubled to 8 KB. Furthermore, there are 54 digital I/O pins, 16 analog inputs and 4 UARTS,
which make the MEGA 2560 more versatile for additional sensors and other applications.
An updated version of the sensor head was developed that used mechanical and chemical bonding to improve the water seal [47].
We also undertook extensive calibration of the turbidity sensor against a Hach turbidimeter based on the ISO 7027 nephelometric
standard for measuring turbidity. We used Formazin (industry turbidity standard) calibration samples that approximated several

10
J. Trevathan et al. Internet of Things 16 (2021) 100429

known NTU values (i.e., <0.2, 20, 200, 1,000, 4,000, and 7,500).
Figure 7 A illustrates a typical response curve for our turbidity sensors. Here, we uniformly input all the pulse width modulation
(PWM) values between 0 and 255 to observe the voltage received (referred to as a PWM sweep). In this instance, we used tap water
(NTU < 1).
All sensors have a curve similar to the one shown in Figure 7 A. However, each sensor varies in its power and response. In this
example, at PWM values >150, the sensor tended to saturate. PWM values <60 were difficult for the sensor to get any sort of reading.
Therefore, the measurable range for the sensor appeared to be PWM values between 60 and 150. The optimal PWM value to use should
be near the top of the curve before the saturation effect occurs. However, the exact PWM value for each sensor is different depending on
the sensor’s response to differing turbidity values.
To establish the optimal PWM value for a sensor we tested 10 sensors on Formazin samples in controlled lighting/temperature
conditions. Once the PWM value was chosen, the calibration curve was determined. Figure 7 B shows the calibration curve for a typical
sensor. The fit for NTU values >100 is good, indicating that the sensor was extremely reliable and accurate for measuring higher levels
of turbidity (100 NTU to 4,000 NTU). However, the reliability and accuracy decreased somewhat when the samples had turbidity
below 100 NTU. This is below the sensitivity range of the sensor. (See [47] for a complete discussion on calibration.)
Ten buoys were constructed. These buoys were deployed on an ongoing basis throughout 2019 and 2020 in Wyaralong Dam
(Figure 6 D), Hinze Dam, Ross River, Slacks Creek and the Burdekin River. Table 7 shows the outcomes for these deployments in terms
of the duration and whether any of the buoys became challenged or failed. The platform formed the basis for an environmental
informatics challenge involving 50 high school students from ten schools. The challenge was undertaken as part of an Seqwater5 Water
for Life Community Grant to engage the students in water quality education and conservation.

4. Performance Considerations

This section provides a performance comparison of the proposed aquatic environmental monitoring platforms in terms of capa­
bilities, stability, cost and flexibility for other applications.

4.1. General System Comparison

Table 8 presents a comparison of the Cave Pearl Data Logger [23] and the three system versions presented in this paper. The Cave
Pearl Data Logger is limited to data logging only (i.e., does not provide telemetry) and does not have a renewable energy supply (i.e.,
no solar charging).
The criteria “stable over time” refers to the stability of the platform against crashes, negative power budget or unknown conditions,
that may cause the platform to become unresponsive. For example, version two of the buoy had an RTC/I2C timing mismatch issue that
resulted in the system potentially going into an infinite sleep. We found that this most commonly occurred if the sensor head leaked
and caused electrical interference. The GPSB proved to be the most reliable with its hardware-controlled timing and multiple failsafe
mechanisms to detect duty cycle issues. We cannot confirm the stability claims for the Cave Pearl Data Loggers.
The overall price point for each of the platforms remained low ($70 – $400 AUD). This represents the material cost only (i.e., the

Figure 7. A) Example electrical response of the sensor to different input pulse width modulation values. B) Voltage versus Nephelometric Turbidity
Units (NTU): Least squares fit.

5
seqwater.com.au

11
J. Trevathan et al. Internet of Things 16 (2021) 100429

Table 7
Third version remote environmental monitoring platform deployment outcomes.
Buoy Id Location Deployment Duration Failed / Challenged Reason

Buoy 1 Slacks Creek 56 days


Buoy 2 Slacks Creek 25 days Vandalised
Buoy 3 Slacks Creek 56 days Lack of sufficient solar power
Buoy 4 Wyaralong Dam 70 days Stolen
Buoy 5 Wyaralong Dam 270 days Solar panel water ingress
Buoy 6 Keyatta Lake 120 days
Buoy 7 Fairfield Waters 120 days
Buoy 8 Ross River 130 days
Buoy 9 Ross River 130 days
Buoy 10 Burdekin River 200 days
Buoy 11 Hinze Dam 168 days Submerged due to water level rise

Table 8
General system comparison of the aquatic environmental monitoring platforms.
Version Microcontroller Sensors Telemetry Solar SRAM Stable Over Approximate Material
Power Time Cost ($AUD)

Data Logger Arduino Pro Mini Lux, temperature, pressure No No 2 KB Yes $70 (plus sensors)
[23] 328
1 Arduino Pro Mini Lux, temperature, turbidity Yes Yes 2 KB No $200
328
2 Arduino Pro Mini Lux, temperature, turbidity Yes Yes 2 KB No $300
328
GPSB Arduino MEGA Lux, temperature, turbidity, Yes Yes 4 KB Yes $400
2560 humidity, status

electronics and physical components – not the labour construction cost). The price increase across each version largely reflects the
additional components and materials that went into waterproofing and enhancing the overall system stability as practical challenges
were overcome. The GPSB costs approximately $2,000 AUD in total between material and labour costs (excluding research and
development costs). In contrast, commercial solutions commence at $20,000 – $60,000 for the base platform (excluding sensors).
The GPSB proves to be the most versatile and powerful out of all of the design proposals. Part of this has to do with the use of the
Arduino Mega 2560 microcontroller and enhanced circuit design (with failsafe mechanisms). This design can incorporate any Arduino-
compatible sensor and can be used in other environmental sensing applications over a continuous period of 12 – 24 months (as
demonstrated during practical deployments).

4.2. Energy Considerations

Table 9 contrasts the differing energy capabilities and requirements between the Cave Pearl Data Logger and the three versions of
our aquatic environmental monitoring platform. The table illustrates the battery capacity (in mWh), solar panel capacity (W) and
power consumption. Power consumption is broken down into the energy required during the sleep and awake periods of the duty
cycles respectively. We also show the average energy consumption over a 24-hour period comprised of 96 duty cycles where the system
wakes every 15 minutes to take samples. Note that the awake duty cycle power measurement for the telemetry buoys is the peak power
usage when communicating via the GSM modem (which is the dominating power consumption factor). Finally, the power budget over
time indicates whether the battery charge capacity is:

Table 9
Energy consumption comparison.
Version Battery Solar Sleep Power Consumption Awake Power Overall Power Consumption Power Budget
Capacity Charge (24-Hour Period) Draw (24-Hour Period) Over Time

Data Logger 7,200 mAh N/A 1.25 mWh - 21,600 mA -negative


[23]
1 14,400 – 6 V 3.2 W 1459.24 mWh 200 mW 295 mWh + 1459.24 mWh -negative
16,000 mAh (continual sleep drain)
2 6,800 – 8,000 6V1W RTC battery only 400 mW 576 mWh +positive
mAh
GPSB 14,400 – 6V2W 0.004 mWh 867 mW 602.12 mWh +positive
16,000 mAh

12
J. Trevathan et al. Internet of Things 16 (2021) 100429

• Negative – the power continues to drain quicker than it is charged (if solar is employed); or
• Positive – with ideal solar conditions, the charge rate exceeds the power drain.

The Cave Pearl Data Loggers run on 3 × 1.5 V alkaline AA batteries. When these batteries are completely new, they supply 7,200
mAh in total. Figure 8 presents the authors’ [23] power consumption analysis indicating how the batteries discharge over a typical 9 –
18 month period. In contrast, our aquatic environmental monitoring platforms use recycled/repackaged lithium ion laptop batteries
[25]. The recycled batteries are rated for 3.6 – 4.2 V and provide between 1,800 and 2,000 mAh per cell. We are also recharging these
batteries via solar energy.
Version one used 8 lithium ion batteries and 4 × 0.8 W solar panels. The duty cycling technique involved sleeping the Arduino.
However, the DC-DC convertor and solar charge controller could not be powered down during sleep cycles. As such, the solar charge
rate could not replenish the power draw on the batteries resulting in a negative power budget over time. Furthermore, the solar panels
were not optimally angled toward the sun, which inhibited their ability to charge the batteries. This meant that the system could only
last 5 – 10 weeks in ideal solar charging conditions.
Version two used 4 Lion batteries and 1 × 1 W solar panel. The duty cycling technique employed a J-K flip-flop that is signaled by
the RTC alarm pulse to latch the MOSFET and power up the system. All components could be powered down between duty cycles,
resulting in a positive power budget (provided full sun for solar charging each day). We only deployed the system for a maximum of
three months over autumn/winter with no noticeable power issues.
The GPSB uses 8 Lion batteries and 2 × 1 W solar panels. The duty cycling technique has hardware timer controlled and allowed all
components to be completely powered down during sleep cycles. In less-than-ideal solar charging conditions (i.e., partial shade or no
sun), the system could retain enough charge for two months of uninterrupted operation. At the time of writing, the GPSB has been
deployed in various applications for 24 months.
Figure 9 A displays the GPSB battery voltages from the Wyaralong Dam deployment from mid-winter to the end of spring. The
figure shows the (averaged) fluctuations in the voltages every fifteen minutes with the (averaged) daily voltage trend. There are three
basic features in this graph:

1 The first five days, where the batteries are increasing in voltage from their pre-deployment charge up to full charge;
2 The middle period up to ninety days where the batteries are maintaining a steady voltage, corresponding to the end of winter; and
3 The last stage from ninety days on where the battery voltage is on a slightly increasing trend, corresponding to the spring period.

Figure 9 B displays the voltages over two five-day periods: the July readings in the middle of winter and the November readings in
the middle of the last month of spring. The battery voltage is higher during spring. This is due to the increasing amount of sunlight
available to the batteries in spring. The batteries are not stressed and would have continued operating indefinitely under this duty
cycle. These results indicate that with recycled Li-ion batteries charged to within an optimal capacity 50%-75%, the system can
essentially last an indefinite period out in the field. That is, the batteries will last for potentially 20 years, which is longer than the
actual life of the buoy. Furthermore, the batteries from retrieved buoys can also be recycled and repurposed for other applications. No
other commercial systems or academic proposals are able to make this claim.

4.3. Data Quality

The deployment results demonstrated that the turbidity sensor was reactive to its environment and indicative of change. Figure 10
presents calibrated data from Wyaralong Dam. The plot shows 48 hours of turbidity readings taken over four consecutive night time
periods of a twelve-hour duration. The blue line displays the raw turbidity, while the red line is a running average of the five previous
and post estimates (i.e., the smoothed turbidity estimates averaged over the five previous and post estimates). In conditions where
there were high amounts of suspended sediment, the NTU value for the sensor would read much higher than water bodies with less
sediment. The sensor’s NTU value would rise significantly during heavy rainfall due to an influx of sediment being washed into the
system, which would gradually fall again as the sediment settled.
There was a degree of ambient IR interference during daytime readings. Even though the sensor was shielded from the direct sun,
the sensor picked up refracted light through the water. The night time readings proved to be more stable when this interference was not
present. Future work would be to take a sensor reading without the transmitter turned on to capture the amount of ambient IR and then
subtract this from the official reading to compensate for the interference.
Different water bodies foul at different rates. Some of the deployments were maintained regularly (i.e., the sensor was cleaned),
whereas other locations did not have any maintenance. The impacts of fouling can be seen as the gradual rise in NTU value over time as
it became more difficult to transmit signals through the lenses. This evidenced the need for regular maintenance to ensure optimal
readings. Future work involves studying the fouling rates of differing water bodies and determining a dynamic calibration adjustment
to offset the fouling error. This will allow the sensors to be deployed for longer periods between maintenance cycles.

4.4. Versatility for Use in Other Environmental Monitoring Applications

The GPSB can be adapted other applications besides the aquatic environmental monitoring system presented so far. Any application
requiring a low-power environmental monitoring solution that uses Arduino-compatible sensors can be interfaced with the platform.
Figure 11 A illustrates how the GPSB is used in a remote water height monitoring system. The device is mounted on a bridge (or

13
J. Trevathan et al. Internet of Things 16 (2021) 100429

Figure 8. Cave Pearl Data Logger battery consumption over time [23].

Figure 9. A) Battery voltage for 200 days. Hourly voltages are averaged over the preceding/succeeding 30 minutes (5 readings); Daily voltages are
averaged over the preceding/succeeding 12 hours (97 readings). B) Battery voltage for two 5-day intervals. Hourly voltages are averaged over the
preceding/succeeding 30 minutes (5 readings).

support) facing down towards a water body. Using an ultrasonic sensor, the device sends a sound wave to the surface of the water and
measures the return time. This allows the platform to remotely determine the water height. We have been able to extend the range of
the sensor to > 10 m with an accuracy of 2 cm. The system has been used to monitor suburban creeks and the hydroperiod around
mangroves in the tidal reaches of river catchments.
Figure 11 B demonstrates another application for the GPSB for air dust pollution monitoring. A light attenuation dust sensor is
integrated with the system. In this example, the system was deployed at a new housing estate construction site. The sensor determines
the air quality based on the drop-off of the light signal due to suspended dust particles in the air.

4.5. Design Goal Comparison

Section 2.4 outlined the desirable design goals for a near real-time aquatic environmental monitoring system. Here we briefly
evaluate the Cave Pearl Project data logger and our three system versions against the stated design goals.
Table 10 presents a high-level comparison of the various proposals and whether or not they achieved the stated design goals. The
Cave Pearl Project data logger has a simple, low-power design that is affordable and can incorporate additional sensors. However, the
system does not have a renewable power supply, does not provide telemetry, and there is insufficient evidence regarding its reliability,
sensor data accuracy and flexibility for use in other applications.

14
J. Trevathan et al. Internet of Things 16 (2021) 100429

Figure 10. Calibrated turbidity data from the Wyaralong Dam deployment.

Figure 11. Water level sensing and air dust sensing.

Table 10
Comparison of the aquatic environmental monitoring platforms against the design goals.
Version Simple Low Renewable Power Plug and Near Real- Accurate Sensor Stable/ Affordable Flexible
Design Power Supply Play Time Data Reliable

Data Logger
[23]
1
2
GPSB

15
J. Trevathan et al. Internet of Things 16 (2021) 100429

Version one of our proposals is successful in achieving a simple design, low-power (albeit with constant vampire power draw while
sleeping), uses solar power, provides telemetry, can incorporate sensors from various vendors, and is low-cost. However, this version
has un-calibrated sensors, degraded over time, and was designed specifically for one application.
Version two addressed the power issues of version one, simplified the design and is a more elegant solution. However, the
manufacturing/parts cost increased slightly and new electrical stability and water ingress issues were encountered which detracted
from the system’s robustness. Furthermore, sensor calibration was largely overlooked during the deployment of this version.
Version three scores positively across all of the design criteria. This is the only one of the proposals to undergo rigorous calibration
and quality assurance of the sensor data. This version is also used throughout multiple deployments in different environmental
conditions over a significant period of time. This version can be used for other environmental monitoring applications such as water
height and air dust sensing (Section 4.4).

5. Conclusions

Historical aquatic environmental observation initiatives are expensive. Most designs are complicated, proprietary, have high power
demands and are not adaptable for other applications. The Cave Pearl Project showed how simple environmental data loggers could be
constructed using inexpensive and readily available open source components. However, the Cave Pearl Project does not provide near
real-time telemetry, renewable power or plug and play simplicity.
This paper presented an IoT platform for affordable aquatic environmental sensing. We proposed the system goals and criteria to be
simplicity of design, low-power consumption, renewable power supply, plug and play, near real-time delivery of reliable and accurate
sensor data, system stability/reliability over time, affordability, and flexibility to be used in other applications. This paper showed how
the system evolved over time to become more robust under increasing functional and field demands to the point where future iterations
could focus on increasing the number of sensor parameters and improving sensor data quality.
The first version of the IoT monitoring platform uses an Arduino Pro Mini microcontroller, Arduino-compatible sensors, solar
power, recycled laptop batteries, telemetry and 3D printed components. This system communicated via the 2G GSM network and
lasted up to five weeks.
The second version of the IoT platform addressed the first version’s shortcomings. This version used a custom-built PCB to mount
the electronics components and a special sensor head PCB for the underwater sensors. This system communicated over the 3G GSM
network and incorporated a new duty cycling technique whereby power is cut to all electronics components when the system is
sleeping. Several buoys were deployed for up to three months and exhibited no power issues.
The third version of the IoT platform (i.e., the GPSB) uses an Arduino Mega 2560, which expanded its computational capacity
dramatically (at the cost of a slight increase in power consumption). The platform uses a highly stable duty cycling mechanism to
achieve low-power consumption that has multiple fail-safes to ensure system stability. The GPSB can accommodate numerous sensors,
which makes it flexible for other applications. A new turbidity sensor design was incorporated into the system and extensive cali­
bration procedures were conducted. Ten buoys were deployed in various water bodies (creeks, lakes, rivers and dam catchments) over
an extended period of time. This version achieved all of the stated desirable design goals.
A performance comparison of all three IoT platform versions and the Cave Pearl logger was presented. Results indicated that the
GPSB is stable and can be deployed on a continuous basis for 18-24 months. This can be extended if more robust glass solar panels are
used. With the current battery pack (recycled lithium ion), the GPSB can sustain long-term operation in ideal solar charging conditions,
and two months operation in the absence of solar charge. Additionally, the lifetime of the recycled batteries can outlast the lifetime of
the buoys. The GPSB is adaptable to other environmental monitoring applications (i.e., water level height monitoring and dust
sensing). Our proposal is novel compared to existing research and commercial systems in terms of its approach to manufacturing (via
social enterprise and recycled batteries), affordability and niche market segment.
Future work involves refining the electrical subsystems, physical design and the definition of a standard for a cleaning mechanism.
Additionally, we intend on incorporating additional sensor parameters from Atlas Scientific such as dissolved oxygen, pH and electrical
conductivity (i.e., salinity) and incorporating machine-learning algorithms to reason over the data. In Section 3.1 we stated a 1.04%
error rate due to transmission issues (mostly due to failing to establish a GSM connection). This can be resolved by utilising the
Arduino’s 2KB EEPROM and the TinySine’s 20MB flash memory storage (or alternately adding a flash chip to the circuitry) to buffer
readings and retransmit in a later duty cycle. Finally, we intend on outlining the software, data management and sensor calibration
aspects of the system – in addition to allowing the user more flexibility in controlling the duty cycling interval remotely or based on low
voltage conditions.

Declaration of competing interest

None

Acknowledgments

This work was supported in part by the Australian Research Council Linkage (LP190101083), Logan City Council EnviroGrants
scheme, Seqwater Community Grants and Griffith University Institute for Integrated and Intelligent Systems. We would like to thank
Ian Trevathan, Ron Johnstone, Jody Kruger and Tom Stevens.

16
J. Trevathan et al. Internet of Things 16 (2021) 100429

References

[1] U.S. Geological Survey, How Much Water is there on Earth, 2020. Last Accessed 01/06/2020.
[2] World Health Organization, Guidelines for drinking-water quality (Vol. 1), World Health Organization, 2004.
[3] R.S. Ayers, D.W. Westcot, Water quality for agriculture (Vol. 29), Food and Agriculture Organization of the United Nations, Rome, 1985.
[4] D.S. Ahearn, R.W. Sheibley, R.A. Dahlgren, M. Anderson, J. Johnson, K.W. Tate, Land use and land cover influence on water quality in the last free-flowing river
draining the western Sierra Nevada, California, Journal of Hydrology 313 (3-4) (2005) 234–247.
[5] P. Kearl, N. Korte, M. Stites, J. Baker, Field comparison of micropurging vs. traditional ground water sampling, Ground Water Monitoring and Remediation 14
(4) (1994) 183–190.
[6] J.D. Plummer, S.C. Long, Monitoring source water for microbial contamination: evaluation of water quality measures, Water research 41 (16) (2007)
3716–3728.
[7] B. Schaffelke, J. Carleton, M. Skuza, I. Zagorskis, M.J. Furnas, Water quality in the inshore Great Barrier Reef lagoon: Implications for long-term monitoring and
management, Marine Pollution Bulletin 65 (4-9) (2012) 249–260.
[8] J. Trevathan, I. Atkinson, W. Read, Establishing low-cost aquatic monitoring networks for developing countries, Communications: Wireless in Developing
Countries and Networks of the Future (2010) 39–50.
[9] C. Peijiang, J. Xuehua, Design and Implementation of Remote monitoring system based on GSM, IEEE Pacific-Asia workshop on computational intelligence and
industrial application 1 (2008) 678–681.
[10] M.T. Lazarescu, Design of a WSN platform for long-term environmental monitoring for IoT applications, IEEE Journal on Emerging and Selected Topics in
Circuits and Systems 3 (1) (2013) 45–54.
[11] S. Adhya, D. Saha, A. Das, J. Jana, H. Saha, An IoT based smart solar photovoltaic remote monitoring and control unit, IEEE Control, Instrumentation, Energy &
Communication (2016) 432–436.
[12] J. Shah, B. Mishra, IoT-enabled low power environment monitoring system for prediction of PM2. 5, Pervasive and Mobile Computing 67 (2020), 101175.
[13] Z. Wang, Q. Wang, X. Hao, The design of the remote water quality monitoring system based on WSN, in: IEEE 5th International Conference on Wireless
Communications, Networking and Mobile Computing (1-4), 2009.
[14] S.A. Ruberg, R.W. Muzzi, S.B. Brandt, J.C. Lane, T.C. Miller, J.J. Gray, S.A. Constant, E.J. Downing, A Wireless Internet-Based Observatory: The Real-time
Coastal Observation Network (ReCON), IEEE OCEANS (2007) 1–6.
[15] L.A. Seders, C.A. Shea, M.D. Lemmon, P.A. Maurice, J.W. Talley, LakeNet: An integrated sensor network for environmental sensing in Lakes, Environ. Eng. Sci.
24 (2007) 183–191.
[16] T. Voigt, F. Osterlind, N. Finne, N. Tsiftes, Z.T. He, J. Eriksson, A. Dunkels, U. Bamstedt, J. Schiller, K. Hjort, Sensor Networking in Aquatic Environments:
Experiences and New Challenges, in: 32nd IEEE Conference on Local Computer Networks, Dublin, Ireland, 2007, pp. 793–798.
[17] T.R. Consi, G. Anderson, G. Barske, H. Bootsma, T. Hansen, J. Janssen, V. Klump, R. Paddock, D. Szmania, K. Verhein, J.T. Waples, Real time observation of the
thermal bar and spring stratification of Lake Michigan with the GLUCOS coastal observatory, in: IEEE Conference OCEANS, Quebec, Canada, 2008, pp. 1–9.
[18] Z. Guo, F. Hong, H. Feng, P. Chen, X. Yang, M. Jiang, OceanSense: Sensor Network of Realtime Ocean Environmental Data Observation and Its Development
Platform, 3rd ACM International Workshop on UnderWater Networks (2008) 101–105.
[19] G.P. Timms, J.W. McCulloch, P. McCarthy, B. Howell, M.D. Dunbabin, K. Hartmann, The tasmanian marine analysis network (TasMAN), IEEE OCEANS (2009)
1–6.
[20] C. Albaladejo, P. Sánchez, A. Iborra, F. Soto, J.A. López, R. Torres, Wireless sensor networks for oceanographic monitoring: A systematic review, Sensors 10 (7)
(2010) 6948–6968.
[21] K. Hill, T. Moltmann, R. Proctor, S. Allen, The Australian Integrated Marine Observing System: delivering data streams to address national and international
research priorities, Marine Technology Society Journal 44 (6) (2010) 65–72.
[22] S. Bainbridge, C. Steinberg, M. Furnas, GBROOS–an ocean observing system for the Great Barrier Reef, International Coral Reef Symposium (2010) 529–533.
[23] P.A. Beddows, E.K. Mallon, Cave pearl data logger: A flexible Arduino-based logging platform for long-Term monitoring in harsh environments, Sensors 18 (2)
(2018) 530.
[24] S. Adam, The emergence of social enterprise (Vol. 4), Psychology Press, 2004.
[25] J. Trevathan, T. Sharp, Up-Cycling E-Waste into Innovative Products Through Social Enterprise, in: 9th International Conference on Smart Cities and Green ICT
Systems (SMARTGREENS), 2020, pp. 185–193.
[26] Defourny, J., & Nyssens, M. (2006). Defining social enterprise. Social enterprise: At the crossroads of market, public policies and civil society, 7, 3-27.
[27] J. Trevathan, R. Johnstone, T. Chiffings, I. Atkinson, N. Bergmann, W. Read, S. Theiss, T. Stevens, SEMAT—the next generation of inexpensive marine
environmental monitoring and measurement systems, Sensors 12 (7) (2012) 9711–9748.
[28] J. Trevathan, R. Johnstone, Smart Environmental Monitoring and Assessment Technologies (SEMAT)—A New Paradigm for Low-Cost, Remote Aquatic
Environmental Monitoring, Sensors 18 (7) (2018) 2248.
[29] J.M. Sadler, D.P. Ames, R. Khattar, A recipe for standards-based data sharing using open source software and low-cost electronics, Journal of Hydroinformatics
(2015), jh2015092.
[30] F.J. Velez, A. Nadziejko, A.L. Christensen, S. Oliveira, T. Rodrigues, V. Costa, J. Gomes, Wireless sensor and networking technologies for swarms of aquatic
surface drones, in: IEEE 82nd Vehicular Technology Conference, 2015, pp. 1–2.
[31] G. Lockridge, B. Dzwonkowski, R. Nelson, S. Powers, Development of a low-cost Arduino-based sonde for coastal applications, Sensors 16 (4) (2016) 528.
[32] C.A. Pérez, F.S. Valles, R.T. Sánchez, M.J. Buendía, F. López-Castejón, J.G. Cervera, Design and deployment of a wireless sensor network for the mar menor
coastal observation system, IEEE Journal of Oceanic Engineering 42 (4) (2017) 966–976.
[33] N. Gunawardena, E.R. Pardyjak, R. Stoll, A. Khadka, Development and evaluation of an open-source, low-cost distributed sensor network for environmental
monitoring applications, Measurement Science and Technology 29 (2) (2018), 024008.
[34] A.T. Demetillo, M.V. Japitana, E.B. Taboada, A system for monitoring water quality in a large aquatic area using wireless sensor network technology,
Sustainable Environment Research 29 (1) (2019) 1–9.
[35] A.D. Wickert, C.T. Sandell, B. Schulz, G.H.C. Ng, Open-source Arduino-compatible data loggers designed for field research, Hydrology & Earth System Sciences
23 (4) (2019).
[36] J.S. Horsburgh, J. Caraballo, M. Ramírez, A.K. Aufdenkampe, D.B. Arscott, S.G. Damiano, Low-cost, open-source, and low-power: But what to do with the data?
Frontiers in Earth Science 7 (67) (2019) 1.
[37] M.S. Hadi, M.R. Maulana, M.A. Mizar, I.A.E. Zaeni, A.N. Afandi, M. Irvan, Self Energy Management System for Battery Operated Data Logger Device Based on
IoT, IEEE Electrical, Electronics and Information Engineering 6 (2019) 133–138.
[38] Q.M. Abdelal, M.A. Alsmadi, N.A. Jaber, Low Cost Monitoring Systems for Environmental and Water Resources Applications, World Environmental and Water
Resources Congress: Emerging and Innovative Technologies and International Perspectives (2019) 96–103.
[39] A.C. Domínguez-Brito, J. Cabrera-Gámez, M. Viera-Pérez, E. Rodríguez-Barrera, L. Hernández-Calvento, A DIY Low-Cost Wireless Wind Data Acquisition System
Used to Study an Arid Coastal Foredune, Sensors 20 (4) (2020) 1064.
[40] F. Xi, Y. Pang, G. Liu, S. Wang, W. Li, C. Zhang, Z.L. Wang, Self-powered intelligent buoy system by water wave energy for sustainable and autonomous wireless
sensing and data transmission, Nano Energy 61 (2019) 1–9.
[41] Z. Zhang, S. Cao, Y. Wang, A long-range 2.4 g network system and scheduling scheme for aquatic environmental monitoring, Electronics 8 (8) (2019) 909.
[42] A. Przybysz, C.M. Duarte, N.R. Geraldi, J. Kosel, M.L. Berumen, Cellular network Marine Sensor Buoy, IEEE Sensors Applications Symposium (2020) 1–6.
[43] S. Guan, J. Wang, C. Jiang, R. Duan, Y. Ren, T.Q. Quek, MagicNet: The Maritime Giant Cellular Network, IEEE Communications Magazine 59 (3) (2021)
117–123.
[44] F. Edition, Guidelines for drinking-water quality, WHO chronicle 38 (4) (2011) 104–108.

17
J. Trevathan et al. Internet of Things 16 (2021) 100429

[45] L. Fewtrell, J. Bartram, Water quality: guidelines, standards & health, IWA publishing, 2001.
[46] S.P. Gorde, M.V. Jadhav, Assessment of water quality parameters: a review, J Eng Res Appl 3 (6) (2013) 2029–2035.
[47] J. Trevathan, W. Read, S. Schmidtke, Towards the Development of an Affordable and Practical Light Attenuation Turbidity Sensor for Remote Near Real-Time
Aquatic Monitoring, Sensors 20 (7) (2020) 1993.

18

You might also like