Murphy 2015

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Talanta 132 (2015) 520–527

Contents lists available at ScienceDirect

Talanta
journal homepage: www.elsevier.com/locate/talanta

A low-cost autonomous optical sensor for water quality monitoring


Kevin Murphy a, Brendan Heery a, Timothy Sullivan a, Dian Zhang a,b, Lizandra Paludetti a,c,
King Tong Lau b, Dermot Diamond b, Ernane Costa c, Noel O'Connor b, Fiona Regan a,n
a
Marine and Environmental Sensing Technology Hub (MESTECH), NCSR, Dublin City University (DCU), Glasnevin, Dublin 9, Ireland
b
CLARITY, Centre for Sensor Web Technologies, NCSR, Dublin City University (DCU), Glasnevin, Dublin 9, Ireland
c
LAFAC Applied and Computational Physics Laboratory, University of São Paulo, Pirassununga, São Paulo 13635-900, Brazil

art ic l e i nf o a b s t r a c t

Article history: A low-cost optical sensor for monitoring the aquatic environment is presented, with the construction
Received 27 May 2014 and design described in detail. The autonomous optical sensor is devised to be environmentally robust,
Received in revised form easily deployable and simple to operate. It consists of a multi-wavelength light source with two
16 September 2014
photodiode detectors capable of measuring the transmission and side-scattering of the light in the
Accepted 18 September 2014
Available online 7 October 2014
detector head. This enables the sensor to give qualitative data on the changes in the optical opacity of the
water. Laboratory tests to confirm colour and turbidity-related responses are described and the results
Keywords: given. The autonomous sensor underwent field deployments in an estuarine environment, and the
Optical sensor results presented here show the sensors capacity to detect changes in opacity and colour relating to
Environmental monitoring
potential pollution events. The application of this low-cost optical sensor is in the area of environmental
Low-cost sensors
pollution alerts to support a water monitoring programme, where multiple such sensors could be
Water quality
In-situ sensing deployed as part of a network.
& 2014 Elsevier B.V. All rights reserved.

1. Introduction The use of in-situ sensors, capable of continuously monitoring


chemical and physical parameters, has been increasing in the recent
The monitoring of European water bodies has historically been past and offers a potential solution to some of the challenges outlined
limited but there has been a step change owing to the adoption of above. They can also provide real-time information and contribute to a
the Water Framework Directive (WFD) in Europe [1]. In response to better representation of long-term trends in aquatic environments [6].
this as well as other legislations including the Bathing Water It is now technologically conceivable to envision a network of sensors
Directive [2], the Birds Directive [3] and the Habitats Directive [4], being deployed at key spatial locations, capable of autonomous
the monitoring of water within Europe will increase in coming years. operation in the field for a year or more and providing real-time
Additionally the pressures of climate change, which will lead to alerts for key events [7]. The area of wireless networked sensors is fast
resource scarcity and water quality changes, give a strong scientific becoming one of the most dynamic and important areas of multi-
and economic argument for the expansion of aquatic monitoring. All disciplinary research [8,9]. The communications and data gathering
of these factors will drive the increase in environmental monitoring abilities of the network need to be of a high quality and suit the tasks
by regulatory agencies, although it should be noted that the high cost at hand. The data from remote continuous monitoring of the environ-
associated with physical sample collection and the transportation to, ment can be, and is already being, used for a variety of applications, in
and subsequent analyses in, laboratories is one of the reasons for the addition to environmental protection [10,11]. Despite the promise of
low level of monitoring over the years [5]. The precision and accuracy such a system many challenges remain using currently available
of these techniques, along with their adherence to standard techni- technology, including a limitation in the analytes measurable, inter-
ques for legal and regulatory procedures has maintained the need for ferences, bio-fouling issues, expense, power requirements and the
these methodologies. The initiative for more monitoring has not need for frequent calibrations. In essence the major issue for environ-
necessarily been matched by an increase in funding for these mental sensing remains, in some instances, being able to accurately
activities, which means that new technology should be employed measure and detect environmental pollutants in laboratory conditions,
to meet the desired increase in spatial and temporal monitoring. but struggling to reproduce these results in the field with continuous
in-situ monitors [12].
If these networks of sensors are to become not only a reality
n
Corresponding author. but commonplace it is necessary to produce reliable, inexpensive
E-mail address: fiona.regan@dcu.ie (F. Regan). in-situ sensors. This means costs need to be considerably reduced

http://dx.doi.org/10.1016/j.talanta.2014.09.045
0039-9140/& 2014 Elsevier B.V. All rights reserved.
K. Murphy et al. / Talanta 132 (2015) 520–527 521

but also, and as importantly, retain and increase the accuracy of of the low-cost sensor to detect shifts in water colour and opacity
the measurements. Some of the recent developments in the this in both the laboratory tests and field trials.
field have included in-situ fluorometers for the measurement of
phytoplankton [13], chemical sensors for pH measurements [14],
and the use of optical refractive index measurements for generic 2. Experimental
water event detection [15]. While some of these sensors are less
precise than laboratory techniques or some existing continuous This section describes how the optical sensor is designed and built,
monitoring sensors they can produce extremely useful data. A as well as outlining all key elements of the sensor. Sections 2.3 and 2.4
barrier to deployment of successful monitoring networks up to now delineate the tests carried out both in the laboratory to characterise
has been cost and availability of analyte specific sensors. The sensor the sensor and the field trials undertaken to observe the performance
described here proposes to address the cost issue – enabling the under real-life conditions.
development of a low-cost network to provide valuable qualitative
information on water quality that can lead to more informed decisions. 2.1. Materials
These devices can be used to inform grab sampling and lead to the
desired quantitative analytical approaches at appropriate times and The photodiodes were manufactured by Texas Instruments
locations. They can give indications of pollution, as early warning (OPT101P) and were purchased from Farnell electronics (www.
systems, or observe trends in environmental conditions. farnell.com), as was the Raspberry Pi (Model B) and its accessories.
This paper presents a low-cost, robust and easily deployable The LEDs were obtained from Radionics via RS-online (www.
sensor for the monitoring of aquatic environments. This multi- rs-online.com) and were manufactured by Avago Technologies
wavelength optical sensor has been designed so that it can be (Amber; HLMP-3850), Kodenshi (IR; OPE5685) and Knightbright
operated on an autonomous basis or in a network of sensors and (Red; L-7113EC, Blue; L-53MBC and Green; L-7113VGCK). The
thus is equipped with telecommunications. The sensor is capable Wixel module was from Pololu Robotics and Electronics and was
of measuring the transmission of light emitted by five separate purchased via Cool Components (www.coolcomponents.co.uk).
LED light sources through water, while simultaneously measuring The PVC-U tubing for the body of the sensor, the stainless steel
the side-scattering of the light measured at right angles to the parts and the copper plating were acquired from local hardware
transmission path. The optical colorimetric sensor (OCS) is devised and marine boating suppliers. The clamps and O-rings were
to give data on bulk water property changes, particularly opacity purchased from Alfa Laval (http://www.alfalaval.com/). The enclo-
and colour changes. sures for the electronics were obtained through Radionics Ireland
The OCS is not a turbidimeter or a chlorophyll sensor, though it and Dexgreen (http://www.dexgreen.com/).
can clearly provide valuable data relating to turbidity events and The food dyes used were E133 Brilliant Blue FCF (blue), E124
primary productivity events. The analytical objectives of the OCS are Ponceau 4R (red), E102 Tartrazine and E142 Green S (green) and
the determination of qualitative variations in water quality based on E110 Sunset Yellow (yellow), from Goodall's of Ireland. The two
opacity or colour changes due to pollution events or temporal turbidity standards were a Formazin solution (246142) from HACH
environmental events. By qualitative it is meant that the optical and a Styrenedivinylbenzene copolymer solution (6073G) from YSI
responses observed using the system are related to opacity changes (a Xylem brand).
in the aquatic environment. This qualitative response is confirmed
by co-deploying and correlating the trends of the OCS with a sensor 2.2. Design and construction
sonde measuring multiple environmental parameters over the test
period. The target water quality parameters are water opacity based The complete sensor system, illustrated in Fig. 1, has the following
on transmission measurements of different coloured LEDs. The features: an LED array light source, two photodiode detectors (901 and
advantages over existing devices include its autonomous nature, 01 to the light source), a low cost (Table 1), a robust re-deployable
low cost and environmental robustness that could lead to multi- design, flexible electronic control, in-built antifouling measures,
sensor deployments for a sensor network. The device can be used to optional GSM communications, an optional integrated temperature
detect pollution events that result in an increase in suspended sensor and a custom built data logger. Table 1 gives a breakdown of
solids, algal blooms or other environmental events. the components cost at the time of manufacture.
Commercial water sensing equipment, in the form of either
multi-parameter sondes or single parameter sensors, dominate the 2.2.1. Sensor body
environmental sensing market. The cost of these sensors can be The OCS is constructed using low cost, robust materials (stainless
prohibitive and when anti-fouling measures and logging systems are steel 308 and 304, PVC-U, copper, rubber and IP 67 and IP 68 rated
taken into the account purchase costs can be further elevated. All enclosures). It consists of a stainless steel sensor head protected by
sensors require cleaning and maintenance at regular intervals by well copper, a foam-filled floatation chamber, double sealed electronics
trained personnel, which makes the cost of ownership sometimes housing, stainless steel mooring point and ballast chain.
prohibitively expensive – particularly during the summer. The OCS Fig. 1(a) gives an exploded view of the device with the numbering
described here would reduce the purchase and maintenance costs of corresponding to Table 2 which lists all of the labelled components
the sensor (including anti-fouling and logging systems). This makes it and materials. Fig. 1(b) presents a schematic of the sensor head
possible to build a sensor network over a large spatial area of an showing the configuration of the LED array, along with relative
interesting aquatic system with a high spatial resolution thereby locations of photodiodes. The OCS can be moored in various ways,
allowing for the creation of a real-time pollution alert system. including standalone (floating), buoy integrated or pier attached. The
This paper is structured as follows; an introduction of the topic OCS is positively buoyant; with the ballast chain attached and
of low-cost optical monitoring in a marine environment is given in connected as a single point mooring, it floats at the surface. In this
this section. The Experimental section gives the details of all the configuration the sensing elements are submerged to a depth of 1 m,
materials and methods used in the construction and testing of the with the electronic and communication housing above the surface.
sensor in the laboratory and in field deployments. The Results and
discussion section analyses and presents the data gathered in the 2.2.2. Sensor head design and Optics
laboratory and in the environment and discusses the outcomes. The sensor head is the part of the sensor which houses the
The Conclusions section reviews the paper, highlighting the ability optics and the detection abilities of the sensor. As the sensor is
522 K. Murphy et al. / Talanta 132 (2015) 520–527

designed to be deployed in marine and other aquatic environ- 45 nm), amber (λPEAK ¼583 nm, FWHM¼36 nm), green (λPEAK ¼
ments, biofilms will most likely form on all components [16]; to 515 nm, FWHM¼ 30 nm) and blue (λPEAK ¼430 nm, FWHM¼60 nm).
minimise this effect the sensor head is constructed of copper The schematic also illustrates the configuration of the various LEDs
which has strong anti-fouling properties [17]. This includes a within the array. Two silicon photodiode detectors (PDs), having a
copper shroud around the light source and detectors and electro- 420% efficiency spectral response ranging from 410 to 1080 nm, are
plating any exposed stainless steel housing with copper. used to detect the light. These photodiodes are placed at 901 and 01
The schematic in Fig. 1(b) shows the sensor head consisting of to the optical path of the light emitted from the LEDs. The pathlength
a light source containing five LEDs of differing wavelengths: between the LED array and the 01 PD is approximately 6 cm, with the
IR (λPEAK ¼ 850 nm, FWHM¼45 nm), red (λPEAK ¼ 627 nm, FWHM¼ 901 PD placed 3 cm from the optical path at right angles to it mid-
way along the path. This enables the transmission of light from
multiple optical wavelengths through the water and the side-
scattering of the light from the water (and from particulate matter
in the water) to be measured simultaneously. The LEDs and the
photodiodes are held in place by an epoxy resin and left exposed to
the environment with no additional interfaces to the water; they are
then encased within a piece of T-shaped copper tubing to prevent
non-scattered light from entering the 901 PD. These specific LEDs
have been chosen as they are readily and inexpensively available,
while providing good coverage across the majority of the visible
spectrum.

2.2.3. Electronics
The OCS system is controlled by a Texas Instruments CC2511F32
micro-controller programmed in a variant of C. This is achieved using
a Wixel, a general-purpose programmable module, mounted on a
board with additional inputs and outputs. In the current configura-
tion the sensor is programmed to read the signals from the two
photodiodes for each LED and the ambient light level (i.e. the signal
level for the photodiodes with all LEDs off) every 10 s. Light levels are
recorded by the PDs connected via trans-impedance amplifiers to the
12 bit analogue to digital convertor of the microcontroller. Individual
Fig. 1. (a) An exploded view of the mechanical components making up the sensor, measurements are produced by averaging a number of readings, to
with the numbering relating to Table 2. (b) Schematic of the sensor head showing reduce electronic noise. The data is outputted via an RS232 and
the orientation of the LED array and photodiodes relative to each other and the
configuration of the five LEDs (green [G], red [R], amber [A], blue [B] and IR [I])
captured by the logger detailed below. The system can be repro-
within the array. (For interpretation of the references to color in this figure legend, grammed to use USB, GSM or low-power RF communications and to
the reader is referred to the web version of this article.) vary the optical sampling rate and sensitivity. A circuit schematic of
the key components is shown in Fig. 2, including the Wixel, the LEDs,
the PDs, the array of amplifiers and the power regulator.
Table 1 The sensor can be powered from any rechargeable, on-board
System component costs at time of manufacture (not including R&D costs). 6 V lead acid battery or using a power cable to connect the mains
via an adaptor. During a measurement cycle the OCS draws
Components Approximate cost (€)
approximately 250 mW of power while communication via GSM
Mechanical 150 can draw up to 2.5 W. The system is programmed to sleep outside
Opto-electronics 250 of the measurement and communication cycles, reducing the
Logger box 150 power consumption to below 1 mW. If the OCS is configured to
Mooring o 100
Total o 650
measure and transmit, using GSM, every minute it can operate off
battery power for approximately 85 h.

Table 2
Mechanical components list.

Part no. (Fig. 1) Component Material Supplier

1 Detector shroud Formed copper plate Hardware store


2 Optical detector Photodiodes in copper tube Farnell Electronics
3 Light source Five LEDs in copper tube Radionics Ireland
4, 9, 14 Sensor housing & clamp Copper plated stainless steel (308) Hardware store
5,6,7,11 Clamp Stainless steel (304) Alfa Laval
8 O-ring seal Rubber Alfa Laval
10,13 Mooring point Stainless steel ring Boating supplier
12 Floatation chamber PVC-U tube, foam filled Hardware store
15 Reinforcing rod Stainless steel (308) Hardware store
16 Reinforcing rod sheath Polythene tube Hardware store
17 Electronics conduit PVC tube Hardware store
18, 19, 21 Electronics housing & clamp Neptune 100 PVC enclosure (IP 68 rated) Dexgreen
20 Electronics enclosure ABS enclosure (IP 67 rated) Radionics Ireland
K. Murphy et al. / Talanta 132 (2015) 520–527 523

Fig. 2. A simplified circuit diagram showing the keys elements of the electronic control system for the sensor.

2.2.4. Data logger


To record the data obtained by the OCS a low-cost prototype data
logger, based on an off-the-shelf single board computer (SBC) called
“Raspberry Pis (R-Pi),” was built. R-Pi consists of a 700 MHz ARM11
processor, Broadcom VideoCore IV GPU, 512 MB RAM, 2  USB
2.0 ports, general purpose input–output (GPIO) interface, RJ-45 LAN
connector and an SD card slot for on-board storage. The board can be
powered by a 5 V DC power supply via a Micro USB or GPIO header.
The power consumption of the R-Pi is 3.5 W maximum depending
on the peripherals attached.
One limitation of the R-Pi is that it does not have a real time
clock (RTC) module on-board; to solve this issue a RasClocks was
purchased and installed on the R-Pi. The cost of the logger and its
housing amounts to under €150, when all parts are purchased
separately without economy of scale.
Fig. 3 shows the structure of the system. A data tunnel between
the logger and OCS is established through a USB to Serial converter.
The USB Wi-Fi access point (AP) allows the operator to download
data onto a mobile device such as a smart phone or laptop wirelessly Fig. 3. Block diagram of the structure of the logger and its communication with the
within a range of up to 45 m, through HTTP protocol. The unit is also sensor and users.
compatible with a GSM/3G dongle, which can send data back to a
data centre in real-time. In addition the data is saved onto the SD
card on-board the logger. The R-Pi is powered by a 5 V external
power source. The AP module (Edimax ew-7711uan) is powered sample. Deionised water was used to dilute the food dyes and the
directly from power supply to increase the system stability. turbidity standards.
In order to test the response of the sensor to different colours,
2.3. Laboratory characteristics of the OCS food dyes were used in various concentrations diluted in water.
Food dyes were chosen to mimic the colour effects of impurities in
Communication between the computer and the laboratory test the water in an effective and cost efficient manner. All of the dyes
sensor was established using a terminal program, via a USB port. listed above were extracted using an electronic micro-pipette and
230 V mains electricity was used as the power source via a 12 VDC mixed with water in concentrations of 0, 10, 20, 30, 40, 50, 60 and
convertor. The sensor was positioned in the sample solution and 70 mL of food dye per L of water.
data logged via the PC terminal. All of the laboratory experiments The effect of opacity on the sensor was also investigated. For
were made using a black container in 1.5 L of water, as this covered these experiments turbidity standards were diluted in water and
the head of the sensor. The tests were performed in triplicate, the turbidity of the sample was quantified. Inhomogeneous mixing
collecting the data for 5 min, obtaining 25 measurements for each of the turbidity standard in water necessitated that the turbidity
524 K. Murphy et al. / Talanta 132 (2015) 520–527

Fig. 4. Average attenuation data (n ¼3) recorded for the red (squares), amber (triangles), green (circles), blue (asterisk) and IR (diamonds) LEDs in the green dye solution for
(a) 01 photodiode and (b) 901 photodiode.

level was cross-checked using a commercial turbidimeter (a TURB 3. Results and discussions
430 IR/T from WTW). Five turbidity measurements were taken for
each dilution and an average was calculated to give the final 3.1. Laboratory tests
concentration. The concentrations of the standards prepared were
2.2, 3.1, 4.3, 5.2, 5.9, 8.1, 14.7, 21.4, 42.4 and 54.2 NTU. The results obtained from the laboratory tests on the food dyes
and turbidity standards are presented here. The response of the OCS
to water opacity trends, and hence water quality, can be observed
by evaluating it with known turbidity standards in the laboratory
2.4. Field deployments and subsequently with a turbidity sensor in the field. Fig. 4 shows
the percentage change in the optical signal recorded on the PD for
Two sites in the Dublin vicinity have been used to perform field each LED versus the concentration of colour dye, compared to water,
trials of the OCS; Malahide estuary (lat.: 531 27' 14'', long.:  61 9') with the ambient light subtracted to correct for light pollution and
and Poolbeg marina (lat.: 531 20' 39', long.:  61 13'). These trials dark current effects. All of the proceeding results for the 901 PD
have been carried out beside commercial sensors for turbidity, exclude the blue LED data sets as the signal recorded by the
chlorophyll and other parameters to enable a true comparison of photodiode for this particular LED is extremely weak and subject
results and validation of the OCS. to large variations, producing potentially inaccurate data. The low
In Malahide, north of Dublin city, the deployment took place in light levels of the LED are exacerbated by the fact that the 901 PD
a typical estuarine body characterised by an elongated shallow universally returns a lower signal than the 01 PD due to the right-
channel. The inner estuary is fed with freshwater from a river and angled optical path and the blue LED is spectrally situated where
does not drain at low tides. The outer estuary is mainly influenced the PD response is the weakest, meaning that a much lower
by the Irish Sea, which drains almost completely at low tides. The percentage of the light is detected. It should be noted that the
average water movement is towards the sea as with all estuaries, amber LED for the 901 PD can give a low signal level and thus a
but it is a mixed estuary as the river flow is less dominant than the larger than average variance, but the results are included.
tidal flow. The main pressures affecting the estuarine catchment In Fig. 4 the response of the 01 PD to the increasing concentra-
water quality is the presence of waste water treatment plants and tion of dye for the LEDs is shown to be linear. The R2 value for the
sewer overflows. red, blue, amber and green LEDs is 40.99, with the R2 value for
Poolbeg Marina is in a busy port environment on the lower Liffey the IR LED equal to 0.732. The results for the 901 PD are also quite
estuary, with the deployment site located in an area of less intense linear with the R2 value for the red, amber and green LEDs40.98
ship traffic. The topography of the estuary is heavily modified, being and the R2 value for the IR LED is equal to 0.781. These results
walled for its whole length and undergoing regular dredging. At the highlight that, for the LEDs in the visible range the green is the
site the water depth is approximately 8 m and the width of the most transmitted and the red and blue are the least transmitted,
channel from wall-to-wall is approximately 260 m. The area acts as a when the green dye is used, as expected.
buffer zone for the freshwater input from the river upstream and the The results from laboratory testing of the OCS with turbidity
tidal flow. Anthropogenic effects have an influence also, with the standards are presented in Fig. 5. The results for the green LED is
input of pollutants (run-off, storm drains, sewage treatment dis- depicted in Fig. 5, giving the change in optical signal for the (top)
charges, industrial discharges, port activity and recreational boating) 01 PD and the (bottom) 901 PD. It was found that the performance
and the modification of water flow (upstream dam releases). of the OCS differs in its response to the turbidity standards
The deployments at Poolbeg and Malahide consisted of the OCS compared to the food dyes. Fig. 5 shows that the response of the
and a commercial sonde which were deployed off a pontoon in the 01 PD to turbidity standards was similar to that of the food dyes.
marina. The commercial system was a YSI sonde (V4 6600) and The other colour LEDs were also found to be transmitting less light
measured temperature, conductivity, pH, depth, chlorophyll, turbidity, to the PD with increasing concentration of turbidity standard. The
dissolved oxygen and blue-green algae concentration. The OCS sensor results for the 901 PD, however, show a different reaction with a
and the YSI system were deployed at a depth of approximately 1 m. general growth in optical signal recorded for increasing
K. Murphy et al. / Talanta 132 (2015) 520–527 525

Fig. 5. Average signal response (n¼3) of the green LED for varying concentrations of turbidity standards for both the 01 (diamonds) and 901 (squares) photodiodes.

concentration of turbidity standard. This figure shows that the systems which provides the capability of communication with a
response of green LED has a linear relationship with increasing wide range of USB enabled or TCP/IP based devices and support for
turbidity standards for both the 01 (R2 ¼0.9986) and 901 many programming languages such as C, Cþ þ , Java, and Python.
(R2 ¼ 0.9939) PDs. This trend is evident also in the other colour The initial cost of R-Pi (€35) is higher than that of the Arduinos
LED results, although not in as significant a manner as that of the (€23–Arduino Due) but add-on devices for internal sensor commu-
green LED. nications and external data transmission are much cheaper, reducing
The increase in signal in the 901 PD for turbidity (compared with the overall cost.
a decrease for the dyes) is as expected and it indicates that the light is A drawback of the R-Pi is that it consumes more power (typically
being scattered off particulate matter in the water and some of the 700 mA) than that of some other embedded systems (as low as
side-scattered portion is subsequently detected by the 901 PD. It 100 mA) but due to its high processing capability only one logger
should be noted that at very low levels of turbidity (o5 NTU) the unit would be required for a medium–large scale sensor network,
results become more unstable with standard deviations of greater lowering the overall power usage. The Raspberry PI also provides
than 5% being observed. Turbidity is used as a surrogate variable for high computation capability which allows for raw sensor data to be
suspended solids concentration [18] and the levels of both turbidity processed on board. Anomaly detection and event detection could
and suspended solids levels are important drivers of population, be performed on-site and alert messages would be sent to the base
community and ecosystem level dynamics of phytoplankton and station only once outliers or abnormal events are detected. The raw
bacterioplankton in estuarine systems [19]. Sustained and sporadic sensor data could be stored on site and would not need to be
increases in turbidity levels are associated with fluctuations in transmitted to the base station. Using the two-way communication
microbial populations and concentration of re-suspended contami- capability, the operator can query the raw data associated with the
nants such as heavy metals and other pollutants [20]. In a number of events detected; such a system would reduce the communication
estuarine systems, strong correlations have been found between (data transmissions) costs which generally consume the majority of
suspended particle concentration and the number of attached the power in a large scale wireless sensor network.
bacteria, an important parameter in estimating microbial population The performance of the sensor in an environmental setting
of bacteria such as E. coli and faecal coliforms [21]. Re-suspension of provided further evidence of the ability of the OCS to measure
benthic sediment leading to higher turbidity levels can create colour and opacity changes in the water. The physical set-up of the
elevated E. coli concentrations in estuarine waters. light source and the detectors allows for various potential methods
of determining turbidity and chlorophyll content, including spec-
3.2. Field deployments tral bands [24,25] and the ratios of the transmitted to side-
scattered light detected by the PDs. In the laboratory studies it
During the deployment periods of the sensor in the field it was was found that the green LED gave the strongest signal and most
found that the system performed successfully in terms of measur- significant response for water opacity and colour changes. There-
ing, acquiring and logging the data. The data logger recorded all fore the results presented relate to the green LED data, using two
the data onto the SD card while allowing it to be accessed also via straightforward methods described below.
Wi-Fi. Although other embedded systems, such as Arduinos, are In Fig. 6, the turbidity recorded by the YSI sonde is shown along
widely used in marine sensor networks [22,23] the R-Pi provides with a scaled version of the optical signal response recorded by the
much more functionalities than these traditional embedded systems. 01 PD for the green LED on the OCS. This illustrates clearly the
These include the supporting of Linux kernel-based operating inverse relationship between a rise in turbidity and a fall in optical
526 K. Murphy et al. / Talanta 132 (2015) 520–527

Fig. 6. Turbidity measured at Poolbeg marina by the YSI sonde (solid blue line) over a 26 h period, with the response of the 01 PD to the green led also plotted (dashed line). This
response has been scaled to show it on the same plot. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 7. Turbidity (dotted blue line) and scaled chlorophyll (dashed red line) measured at Malahide by the YSI sonde over a 14.5 h period, along with optical data from the
green LED (solid black line). The method for calculating the optical data is given above. (For interpretation of the references to color in this figure legend, the reader is
referred to the web version of this article.)

signal for the 01 PD, which is in agreement with the results obtained where GLED is the green LED result, G0 and G90 are the optical
in the laboratory tests depicted in Fig. 5. The signal response has signals for the green LED from the 01 and 901 PDs respectively,
been corrected for ambient light and has been smoothed using a D0 and D90 are the ambient light signals from the 01 and 901 PDs
moving average filter. The optical signal response in Fig. 6 has been and S is a scaling factor.
scaled and thus has arbitrary units. The correlation between the 01 Performing this has the effect of smoothing out the event driven
PD response and the YSI measured turbidity was calculated using spikes in turbidity, dominated by particulate matter, from the
the Pearson method to be  0.8322. optical data, leaving a more stable signal; the residual optical data
The results shown in Fig. 7 illustrate how it may be possible to tracks the changes in the chlorophyll concentration. A question
do the same for the assessment of colour presence in the water; remains as to whether the parameter being represented by the
here chlorophyll is used as an example. In Fig. 7 the turbidity and a optical data will always be a reliable indicator of chlorophyll. It is
scaled set of chlorophyll concentrations measured by the YSI clear from the measurements that there is a relationship between
sonde are given, along with optical data from the OCS. This data this optical dataset and a surrogate chlorophyll value which gives an
was recorded during the Malahide deployment and the section of indicator of chlorophyll occurrence and variation. A Pearson corre-
the data displayed covers approximately 14.5 h. The optical data lation carried out on the residual optical data and the chlorophyll
shown is for the green LED. The raw signal has been adjusted using concentration measured by the YSI sonde gives a value of 0.8123.
the following calculation: Based on the results to date it is believed that this sensor could
  be employed, within a network of sensors, to identify event changes
G90  D90 in the bulk water quality parameters, such as variations in water
GLED ¼  S;
G0  D0 clarity and primary productivity-related turbidity. The low-cost
K. Murphy et al. / Talanta 132 (2015) 520–527 527

nature of the OCS means that multiple sensors could be deployed Also, QUESTOR Research Centre under Grant DCU8/10/2013 and
across a large area, with a high spatial density. It is envisaged that a Science Foundation Ireland under grant 07/CE/I1147.
small number of auto-samplers and/or more expensive, sophisti- The authors also wish to thank the staff at Poolbeg and Malahide
cated sensors could be deployed in tandem with multiple low-cost Marinas for permission, assistance and access to the deployment
OCS units to ground truth the data and verify the bulk water location and facilities. Some students have assisted in this work and
property changes. Alternatively the events detected by the low-cost deserve thanks, Matthew Meagher for producing drawings and
sensor network could inform and drive the location and timing of Camilla Nardi Pinto for testing and calibration.
manual grab sampling for laboratory analysis.

4. Conclusions References

The design, building and operation of a low-cost, multi-wavelength [1] European Water Framework Directive, Directive 2000/60/EC.
optical system for water quality monitoring has been illustrated and [2] European Bathing Water Directive, Directive 2006/7/EC.
[3] European Birds Directive, Directive 2009/147/EC.
described. The OCS is capable of measuring the change in optical [4] European Habitats Directive, Directive 1992/43/EC.
signal along two different optical paths (transmitted and side- [5] D. Diamond, K.T. Lau, S. Brady, J. Cleary, Talanta 75 (3) (2008) 606–612. http:
scattered). The performance of the system has been tested in the //dx.doi.org/10.1016/j.talanta.2007.11.022.
[6] R. Greenwood, J. Webster, F. Regan, Royal Society of Chemistry Report 2007.
laboratory with food dye and turbidity standard solutions of varying [7] J. Grath, R. Ward, A. Scheidleder, P. Quevauviller, J. Environ. Monit. 9 (11)
concentrations. These results show strong linear correlations between (2007) 1162–1175. http://dx.doi.org/10.1039/B710665N.
the optical signal response and turbidity concentration, as well as [8] C. Albaladejo, P. Sanchez, A. Iborra, F. Soto, J.A. Lopez, R. Torres, Sensors 10 (7)
colour change from the food dyes. The system was deployed in the (2010) 6948–6968. http://dx.doi.org/10.3390/s100706948.
[9] P.W. Rundel, E.A. Graham, M.F. Allen, J.C. Fisher, T.C. Harmon, New Phytol. 182
field at different locations and the recorded data correlated well with (3) (2009) 589–607. http://dx.doi.org/10.1111/j.1469-8137.2009.02811.x.
commercial water quality sensors. The results from the field trials [10] S.M. Glenn, T.D. Dickey, B. Parker, W. Boicourt, Oceanography 13 (1) (2000)
highlight that the system has the ability to detect both sudden and 24–34. http://dx.doi.org/10.5670/oceanog.2000.50.
[11] T.S. Moore, K.M. Mullaugh, R.R. Holyoke, A.S. Madison, M. Yucel, G.W. Luther
significant changes in water opacity arising from environmental
III, Annu. Rev. Mar. Sci. 1 (2009) 91–115. http://dx.doi.org/10.1146/annurev.
events. marine.010908.163817.
The work presented in this paper has outlined the first genera- [12] W. Bourgeois, A.C. Romain, J. Nicolas, R.M. Stuetz, J. Environ. Monit. 5 (6)
tion of the sensor development and results. Further work and (2003) 852–860. http://dx.doi.org/10.1039/B307905H.
[13] T. Leeuw, E.S. Boss, D.L. Wright, Sensors 13 (6) (2013) 7872–7883. http://dx.
potential adaptations are required to produce a sensor that can be doi.org/10.3390/s130607872.
easily and quickly calibrated in the field; along with more sophis- [14] B.H. Tangena, et. al., Water Contamination Emergencies (2011), pp. 13–31 .
ticated data analysis, this will enable the sensor to produce the most 10.1039/9781849733199-00013
[15] A. Radu, S. Anastasova, C. Fay, D. Diamond, J. Bobacka, A. Lewenstam, Proceedings of
desirable results from the data collected. In order to facilitate this,
the IEEE Sensors Conference, 2010, pp. 1487–1490. 10.1109/ICSENS.2010.5690357
additional longer term deployments in test locations are planned. It [16] A. Whelan, F. Regan., J. Environ. Monit.8 (9) (2006) 880–886. http://dx.doi.org/
is now possible to gather this type of data over a large spatial area, in 10.1039/B603289C.
real-time, at a reasonable cost; this would make these kinds of [17] L.D. Chambers, et al., Surf. Coat. Technol. 201 (6) (2006) 3642–3652. http://dx.
doi.org/10.1016/j.surfcoat.2006.08.129.
systems particularly attractive to lower income countries struggling [18] W.F. Henley, M.A. Patterson, R.J. Neves, A. Dennis Lemly, Rev. Fish. Sci. 8 (2)
with the effects of climate change and water related challenges. The (2000) 125–139.
intention is to commercialise the OCS as a real-time pollution [19] N.K. Goosen, J. Kromkamp, J. Peene, P. van Rijswijk, P. van Breugel, J. Mar. Sys.
22 (2–3) (1999) 151–171.
monitoring system.
[20] A. Tremblay, L. Varfalvy, C. Roehm, M. Garneau (Eds.), Environmental Science
and Engineering Series, Springer, New York.
[21] Y.A. Pachepsky, D.R. Shelton, Crit. Rev. Environ. Sci. Technol. 41 (12) (2011)
Acknowledgements 1067–1110.
[22] N.W. Bergmann, M. Wallace, E. Calia, Proceedings of the Sixth International
Conference on Intelligent Sensors, Sensor Networks and Information Proces-
The authors acknowledge the funding of this research through the sing (ISSNIP), 2010, pp. 19–24 . 10.1109/ISSNIP.2010.5706802
Beaufort Marine Research Award, carried out under the Sea Change [23] J. Trevathan, et al., Sensors 12 (7) (2012) 9711–9748. http://dx.doi.org/10.3390/
Strategy and the Strategy for Science Technology and Innovation s120709711.
[24] A.A. Gilerson, et al., Opt. Express 18 (23) (2010) 24109–24125. http://dx.doi.
(2006–2013), with the support of the Marine Institute, funded under
org/10.1364/OE.18.024109.
the Marine Research Sub-Programme of the National Development [25] F.E. Hoge, R.N. Swift, Appl. Opt. 25 (15) (1986) 2571–2583. http://dx.doi.org/
Plan 2007–2013 under grant BEAU/SENS/10. 10.1364/AO.25.003677.

You might also like