A low-cost autonomous optical sensor for water quality monitoring

SAVE OFFLINE

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 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 b

art ic l e i nf o

a b s t r a c t

Article history: Received 27 May 2014 Received in revised form 16 September 2014 Accepted 18 September 2014 Available online 7 October 2014

A low-cost optical sensor for monitoring the aquatic environment is presented, with the construction and design described in detail. The autonomous optical sensor is devised to be environmentally robust, easily deployable and simple to operate. It consists of a multi-wavelength light source with two photodiode detectors capable of measuring the transmission and side-scattering of the light in the 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 given. The autonomous sensor underwent field deployments in an estuarine environment, and the results presented here show the sensors capacity to detect changes in opacity and colour relating to potential pollution events. The application of this low-cost optical sensor is in the area of environmental pollution alerts to support a water monitoring programme, where multiple such sensors could be deployed as part of a network. & 2014 Elsevier B.V. All rights reserved.

Keywords: Optical sensor Environmental monitoring Low-cost sensors Water quality In-situ sensing

1. Introduction The monitoring of European water bodies has historically been limited but there has been a step change owing to the adoption of the Water Framework Directive (WFD) in Europe [1]. In response to this as well as other legislations including the Bathing Water Directive [2], the Birds Directive [3] and the Habitats Directive [4], the monitoring of water within Europe will increase in coming years. Additionally the pressures of climate change, which will lead to resource scarcity and water quality changes, give a strong scientific and economic argument for the expansion of aquatic monitoring. All of these factors will drive the increase in environmental monitoring by regulatory agencies, although it should be noted that the high cost associated with physical sample collection and the transportation to, and subsequent analyses in, laboratories is one of the reasons for the low level of monitoring over the years [5]. The precision and accuracy of these techniques, along with their adherence to standard techniques for legal and regulatory procedures has maintained the need for these methodologies. The initiative for more monitoring has not necessarily been matched by an increase in funding for these activities, which means that new technology should be employed to meet the desired increase in spatial and temporal monitoring.

n

Corresponding author. E-mail address: fi[email protected] (F. Regan).

http://dx.doi.org/10.1016/j.talanta.2014.09.045 0039-9140/& 2014 Elsevier B.V. All rights reserved.

The use of in-situ sensors, capable of continuously monitoring chemical and physical parameters, has been increasing in the recent past and offers a potential solution to some of the challenges outlined above. They can also provide real-time information and contribute to a better representation of long-term trends in aquatic environments [6]. It is now technologically conceivable to envision a network of sensors being deployed at key spatial locations, capable of autonomous operation in the field for a year or more and providing real-time alerts for key events [7]. The area of wireless networked sensors is fast becoming one of the most dynamic and important areas of multidisciplinary research [8,9]. The communications and data gathering abilities of the network need to be of a high quality and suit the tasks at hand. The data from remote continuous monitoring of the environment can be, and is already being, used for a variety of applications, in addition to environmental protection [10,11]. Despite the promise of such a system many challenges remain using currently available technology, including a limitation in the analytes measurable, interferences, bio-fouling issues, expense, power requirements and the need for frequent calibrations. In essence the major issue for environmental sensing remains, in some instances, being able to accurately measure and detect environmental pollutants in laboratory conditions, 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 but commonplace it is necessary to produce reliable, inexpensive in-situ sensors. This means costs need to be considerably reduced

K. Murphy et al. / Talanta 132 (2015) 520–527

but also, and as importantly, retain and increase the accuracy of the measurements. Some of the recent developments in the this 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 water event detection [15]. While some of these sensors are less precise than laboratory techniques or some existing continuous monitoring sensors they can produce extremely useful data. A barrier to deployment of successful monitoring networks up to now has been cost and availability of analyte specific sensors. The sensor described here proposes to address the cost issue – enabling the development of a low-cost network to provide valuable qualitative information on water quality that can lead to more informed decisions. These devices can be used to inform grab sampling and lead to the desired quantitative analytical approaches at appropriate times and locations. They can give indications of pollution, as early warning systems, or observe trends in environmental conditions. This paper presents a low-cost, robust and easily deployable sensor for the monitoring of aquatic environments. This multiwavelength optical sensor has been designed so that it can be operated on an autonomous basis or in a network of sensors and thus is equipped with telecommunications. The sensor is capable of measuring the transmission of light emitted by five separate LED light sources through water, while simultaneously measuring the side-scattering of the light measured at right angles to the transmission path. The optical colorimetric sensor (OCS) is devised to give data on bulk water property changes, particularly opacity and colour changes. The OCS is not a turbidimeter or a chlorophyll sensor, though it can clearly provide valuable data relating to turbidity events and primary productivity events. The analytical objectives of the OCS are the determination of qualitative variations in water quality based on opacity or colour changes due to pollution events or temporal environmental events. By qualitative it is meant that the optical responses observed using the system are related to opacity changes in the aquatic environment. This qualitative response is confirmed by co-deploying and correlating the trends of the OCS with a sensor sonde measuring multiple environmental parameters over the test period. The target water quality parameters are water opacity based on transmission measurements of different coloured LEDs. The advantages over existing devices include its autonomous nature, low cost and environmental robustness that could lead to multisensor deployments for a sensor network. The device can be used to detect pollution events that result in an increase in suspended solids, algal blooms or other environmental events. Commercial water sensing equipment, in the form of either multi-parameter sondes or single parameter sensors, dominate the environmental sensing market. The cost of these sensors can be prohibitive and when anti-fouling measures and logging systems are taken into the account purchase costs can be further elevated. All sensors require cleaning and maintenance at regular intervals by well trained personnel, which makes the cost of ownership sometimes prohibitively expensive – particularly during the summer. The OCS described here would reduce the purchase and maintenance costs of the sensor (including anti-fouling and logging systems). This makes it possible to build a sensor network over a large spatial area of an interesting aquatic system with a high spatial resolution thereby allowing for the creation of a real-time pollution alert system. This paper is structured as follows; an introduction of the topic of low-cost optical monitoring in a marine environment is given in this section. The Experimental section gives the details of all the materials and methods used in the construction and testing of the sensor in the laboratory and in field deployments. The Results and discussion section analyses and presents the data gathered in the laboratory and in the environment and discusses the outcomes. The Conclusions section reviews the paper, highlighting the ability

521

of the low-cost sensor to detect shifts in water colour and opacity in both the laboratory tests and field trials. 2. Experimental This section describes how the optical sensor is designed and built, as well as outlining all key elements of the sensor. Sections 2.3 and 2.4 delineate the tests carried out both in the laboratory to characterise the sensor and the field trials undertaken to observe the performance under real-life conditions. 2.1. Materials The photodiodes were manufactured by Texas Instruments (OPT101P) and were purchased from Farnell electronics (www. farnell.com), as was the Raspberry Pi (Model B) and its accessories. The LEDs were obtained from Radionics via RS-online (www. rs-online.com) and were manufactured by Avago Technologies (Amber; HLMP-3850), Kodenshi (IR; OPE5685) and Knightbright (Red; L-7113EC, Blue; L-53MBC and Green; L-7113VGCK). The Wixel module was from Pololu Robotics and Electronics and was purchased via Cool Components (www.coolcomponents.co.uk). The PVC-U tubing for the body of the sensor, the stainless steel parts and the copper plating were acquired from local hardware and marine boating suppliers. The clamps and O-rings were purchased from Alfa Laval (http://www.alfalaval.com/). The enclosures for the electronics were obtained through Radionics Ireland and Dexgreen (http://www.dexgreen.com/). The food dyes used were E133 Brilliant Blue FCF (blue), E124 Ponceau 4R (red), E102 Tartrazine and E142 Green S (green) and E110 Sunset Yellow (yellow), from Goodall's of Ireland. The two turbidity standards were a Formazin solution (246142) from HACH and a Styrenedivinylbenzene copolymer solution (6073G) from YSI (a Xylem brand). 2.2. Design and construction The complete sensor system, illustrated in Fig. 1, has the following features: an LED array light source, two photodiode detectors (901 and 01 to the light source), a low cost (Table 1), a robust re-deployable design, flexible electronic control, in-built antifouling measures, optional GSM communications, an optional integrated temperature sensor and a custom built data logger. Table 1 gives a breakdown of the components cost at the time of manufacture. 2.2.1. Sensor body The OCS is constructed using low cost, robust materials (stainless steel 308 and 304, PVC-U, copper, rubber and IP 67 and IP 68 rated enclosures). It consists of a stainless steel sensor head protected by copper, a foam-filled floatation chamber, double sealed electronics housing, stainless steel mooring point and ballast chain. Fig. 1(a) gives an exploded view of the device with the numbering corresponding to Table 2 which lists all of the labelled components and materials. Fig. 1(b) presents a schematic of the sensor head showing the configuration of the LED array, along with relative locations of photodiodes. The OCS can be moored in various ways, including standalone (floating), buoy integrated or pier attached. The OCS is positively buoyant; with the ballast chain attached and connected as a single point mooring, it floats at the surface. In this configuration the sensing elements are submerged to a depth of 1 m, with the electronic and communication housing above the surface. 2.2.2. Sensor head design and Optics The sensor head is the part of the sensor which houses the 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 environments, biofilms will most likely form on all components [16]; to minimise this effect the sensor head is constructed of copper which has strong anti-fouling properties [17]. This includes a copper shroud around the light source and detectors and electroplating any exposed stainless steel housing with copper. The schematic in Fig. 1(b) shows the sensor head consisting of a light source containing five LEDs of differing wavelengths: IR (λPEAK ¼ 850 nm, FWHM¼45 nm), red (λPEAK ¼ 627 nm, FWHM¼

Fig. 1. (a) An exploded view of the mechanical components making up the sensor, with the numbering relating to Table 2. (b) Schematic of the sensor head showing 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]) within the array. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Table 1 System component costs at time of manufacture (not including R&D costs). Components

Approximate cost (€)

Mechanical Opto-electronics Logger box Mooring Total

150 250 150 o 100 o 650

45 nm), amber (λPEAK ¼583 nm, FWHM¼36 nm), green (λPEAK ¼ 515 nm, FWHM¼ 30 nm) and blue (λPEAK ¼430 nm, FWHM¼60 nm). The schematic also illustrates the configuration of the various LEDs within the array. Two silicon photodiode detectors (PDs), having a 420% efficiency spectral response ranging from 410 to 1080 nm, are used to detect the light. These photodiodes are placed at 901 and 01 to the optical path of the light emitted from the LEDs. The pathlength between the LED array and the 01 PD is approximately 6 cm, with the 901 PD placed 3 cm from the optical path at right angles to it midway along the path. This enables the transmission of light from multiple optical wavelengths through the water and the sidescattering 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 configuration 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 measurements are produced by averaging a number of readings, to reduce electronic noise. The data is outputted via an RS232 and captured by the logger detailed below. The system can be reprogrammed to use USB, GSM or low-power RF communications and to 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. The sensor can be powered from any rechargeable, on-board 6 V lead acid battery or using a power cable to connect the mains via an adaptor. During a measurement cycle the OCS draws approximately 250 mW of power while communication via GSM can draw up to 2.5 W. The system is programmed to sleep outside of the measurement and communication cycles, reducing the power consumption to below 1 mW. If the OCS is configured to 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 2 3 4, 9, 14 5,6,7,11 8 10,13 12 15 16 17 18, 19, 21 20

Detector shroud Optical detector Light source Sensor housing & clamp Clamp O-ring seal Mooring point Floatation chamber Reinforcing rod Reinforcing rod sheath Electronics conduit Electronics housing & clamp Electronics enclosure

Formed copper plate Photodiodes in copper tube Five LEDs in copper tube Copper plated stainless steel (308) Stainless steel (304) Rubber Stainless steel ring PVC-U tube, foam filled Stainless steel (308) Polythene tube PVC tube Neptune 100 PVC enclosure (IP 68 rated) ABS enclosure (IP 67 rated)

Hardware store Farnell Electronics Radionics Ireland Hardware store Alfa Laval Alfa Laval Boating supplier Hardware store Hardware store Hardware store Hardware store Dexgreen 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 within a range of up to 45 m, through HTTP protocol. The unit is also 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 directly from power supply to increase the system stability. 2.3. Laboratory characteristics of the OCS Communication between the computer and the laboratory test sensor was established using a terminal program, via a USB port. 230 V mains electricity was used as the power source via a 12 VDC convertor. The sensor was positioned in the sample solution and data logged via the PC terminal. All of the laboratory experiments were made using a black container in 1.5 L of water, as this covered the head of the sensor. The tests were performed in triplicate, collecting the data for 5 min, obtaining 25 measurements for each

Fig. 3. Block diagram of the structure of the logger and its communication with the sensor and users.

sample. Deionised water was used to dilute the food dyes and the turbidity standards. In order to test the response of the sensor to different colours, food dyes were used in various concentrations diluted in water. Food dyes were chosen to mimic the colour effects of impurities in the water in an effective and cost efficient manner. All of the dyes listed above were extracted using an electronic micro-pipette and mixed with water in concentrations of 0, 10, 20, 30, 40, 50, 60 and 70 mL of food dye per L of water. The effect of opacity on the sensor was also investigated. For these experiments turbidity standards were diluted in water and the turbidity of the sample was quantified. Inhomogeneous mixing 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 430 IR/T from WTW). Five turbidity measurements were taken for each dilution and an average was calculated to give the final 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.

2.4. Field deployments Two sites in the Dublin vicinity have been used to perform field trials of the OCS; Malahide estuary (lat.: 531 27' 14'', long.:  61 9') and Poolbeg marina (lat.: 531 20' 39', long.:  61 13'). These trials have been carried out beside commercial sensors for turbidity, chlorophyll and other parameters to enable a true comparison of results and validation of the OCS. In Malahide, north of Dublin city, the deployment took place in a typical estuarine body characterised by an elongated shallow channel. The inner estuary is fed with freshwater from a river and does not drain at low tides. The outer estuary is mainly influenced by the Irish Sea, which drains almost completely at low tides. The average water movement is towards the sea as with all estuaries, but it is a mixed estuary as the river flow is less dominant than the tidal flow. The main pressures affecting the estuarine catchment water quality is the presence of waste water treatment plants and sewer overflows. Poolbeg Marina is in a busy port environment on the lower Liffey estuary, with the deployment site located in an area of less intense ship traffic. The topography of the estuary is heavily modified, being walled for its whole length and undergoing regular dredging. At the site the water depth is approximately 8 m and the width of the channel from wall-to-wall is approximately 260 m. The area acts as a buffer zone for the freshwater input from the river upstream and the tidal flow. Anthropogenic effects have an influence also, with the input of pollutants (run-off, storm drains, sewage treatment discharges, industrial discharges, port activity and recreational boating) and the modification of water flow (upstream dam releases). The deployments at Poolbeg and Malahide consisted of the OCS and a commercial sonde which were deployed off a pontoon in the marina. The commercial system was a YSI sonde (V4 6600) and measured temperature, conductivity, pH, depth, chlorophyll, turbidity, dissolved oxygen and blue-green algae concentration. The OCS sensor and the YSI system were deployed at a depth of approximately 1 m.

3. Results and discussions 3.1. Laboratory tests 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 and subsequently with a turbidity sensor in the field. Fig. 4 shows the percentage change in the optical signal recorded on the PD for each LED versus the concentration of colour dye, compared to water, with the ambient light subtracted to correct for light pollution and dark current effects. All of the proceeding results for the 901 PD exclude the blue LED data sets as the signal recorded by the photodiode for this particular LED is extremely weak and subject to large variations, producing potentially inaccurate data. The low light levels of the LED are exacerbated by the fact that the 901 PD universally returns a lower signal than the 01 PD due to the rightangled optical path and the blue LED is spectrally situated where the PD response is the weakest, meaning that a much lower percentage of the light is detected. It should be noted that the amber LED for the 901 PD can give a low signal level and thus a larger than average variance, but the results are included. In Fig. 4 the response of the 01 PD to the increasing concentration of dye for the LEDs is shown to be linear. The R2 value for the red, blue, amber and green LEDs is 40.99, with the R2 value for the IR LED equal to 0.732. The results for the 901 PD are also quite linear with the R2 value for the red, amber and green LEDs40.98 and the R2 value for the IR LED is equal to 0.781. These results highlight that, for the LEDs in the visible range the green is the most transmitted and the red and blue are the least transmitted, when the green dye is used, as expected. The results from laboratory testing of the OCS with turbidity standards are presented in Fig. 5. The results for the green LED is depicted in Fig. 5, giving the change in optical signal for the (top) 01 PD and the (bottom) 901 PD. It was found that the performance of the OCS differs in its response to the turbidity standards compared to the food dyes. Fig. 5 shows that the response of the 01 PD to turbidity standards was similar to that of the food dyes. The other colour LEDs were also found to be transmitting less light to the PD with increasing concentration of turbidity standard. The results for the 901 PD, however, show a different reaction with a 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 response of green LED has a linear relationship with increasing turbidity standards for both the 01 (R2 ¼0.9986) and 901 (R2 ¼ 0.9939) PDs. This trend is evident also in the other colour LED results, although not in as significant a manner as that of the green LED. The increase in signal in the 901 PD for turbidity (compared with a decrease for the dyes) is as expected and it indicates that the light is being scattered off particulate matter in the water and some of the side-scattered portion is subsequently detected by the 901 PD. It should be noted that at very low levels of turbidity (o5 NTU) the results become more unstable with standard deviations of greater than 5% being observed. Turbidity is used as a surrogate variable for suspended solids concentration [18] and the levels of both turbidity and suspended solids levels are important drivers of population, community and ecosystem level dynamics of phytoplankton and bacterioplankton in estuarine systems [19]. Sustained and sporadic increases in turbidity levels are associated with fluctuations in microbial populations and concentration of re-suspended contaminants such as heavy metals and other pollutants [20]. In a number of estuarine systems, strong correlations have been found between suspended particle concentration and the number of attached bacteria, an important parameter in estimating microbial population of bacteria such as E. coli and faecal coliforms [21]. Re-suspension of benthic sediment leading to higher turbidity levels can create elevated E. coli concentrations in estuarine waters. 3.2. Field deployments During the deployment periods of the sensor in the field it was found that the system performed successfully in terms of measuring, acquiring and logging the data. The data logger recorded all the data onto the SD card while allowing it to be accessed also via Wi-Fi. Although other embedded systems, such as Arduinos, are widely used in marine sensor networks [22,23] the R-Pi provides much more functionalities than these traditional embedded systems. These include the supporting of Linux kernel-based operating

systems which provides the capability of communication with a wide range of USB enabled or TCP/IP based devices and support for many programming languages such as C, Cþ þ , Java, and Python. The initial cost of R-Pi (€35) is higher than that of the Arduinos (€23–Arduino Due) but add-on devices for internal sensor communications and external data transmission are much cheaper, reducing the overall cost. A drawback of the R-Pi is that it consumes more power (typically 700 mA) than that of some other embedded systems (as low as 100 mA) but due to its high processing capability only one logger unit would be required for a medium–large scale sensor network, lowering the overall power usage. The Raspberry PI also provides high computation capability which allows for raw sensor data to be processed on board. Anomaly detection and event detection could be performed on-site and alert messages would be sent to the base station only once outliers or abnormal events are detected. The raw sensor data could be stored on site and would not need to be transmitted to the base station. Using the two-way communication capability, the operator can query the raw data associated with the events detected; such a system would reduce the communication (data transmissions) costs which generally consume the majority of the power in a large scale wireless sensor network. The performance of the sensor in an environmental setting provided further evidence of the ability of the OCS to measure colour and opacity changes in the water. The physical set-up of the light source and the detectors allows for various potential methods of determining turbidity and chlorophyll content, including spectral bands [24,25] and the ratios of the transmitted to sidescattered light detected by the PDs. In the laboratory studies it was found that the green LED gave the strongest signal and most significant response for water opacity and colour changes. Therefore the results presented relate to the green LED data, using two straightforward methods described below. In Fig. 6, the turbidity recorded by the YSI sonde is shown along with a scaled version of the optical signal response recorded by the 01 PD for the green LED on the OCS. This illustrates clearly the 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 in the laboratory tests depicted in Fig. 5. The signal response has been corrected for ambient light and has been smoothed using a moving average filter. The optical signal response in Fig. 6 has been scaled and thus has arbitrary units. The correlation between the 01 PD response and the YSI measured turbidity was calculated using the Pearson method to be  0.8322. The results shown in Fig. 7 illustrate how it may be possible to do the same for the assessment of colour presence in the water; here chlorophyll is used as an example. In Fig. 7 the turbidity and a scaled set of chlorophyll concentrations measured by the YSI sonde are given, along with optical data from the OCS. This data was recorded during the Malahide deployment and the section of the data displayed covers approximately 14.5 h. The optical data shown is for the green LED. The raw signal has been adjusted using the following calculation: GLED ¼

  G90  D90  S; G0  D0

where GLED is the green LED result, G0 and G90 are the optical signals for the green LED from the 01 and 901 PDs respectively, D0 and D90 are the ambient light signals from the 01 and 901 PDs and S is a scaling factor. Performing this has the effect of smoothing out the event driven spikes in turbidity, dominated by particulate matter, from the optical data, leaving a more stable signal; the residual optical data tracks the changes in the chlorophyll concentration. A question remains as to whether the parameter being represented by the optical data will always be a reliable indicator of chlorophyll. It is clear from the measurements that there is a relationship between this optical dataset and a surrogate chlorophyll value which gives an indicator of chlorophyll occurrence and variation. A Pearson correlation carried out on the residual optical data and the chlorophyll concentration measured by the YSI sonde gives a value of 0.8123. Based on the results to date it is believed that this sensor could be employed, within a network of sensors, to identify event changes in the bulk water quality parameters, such as variations in water 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 across a large area, with a high spatial density. It is envisaged that a small number of auto-samplers and/or more expensive, sophisticated sensors could be deployed in tandem with multiple low-cost OCS units to ground truth the data and verify the bulk water property changes. Alternatively the events detected by the low-cost sensor network could inform and drive the location and timing of manual grab sampling for laboratory analysis.

Also, QUESTOR Research Centre under Grant DCU8/10/2013 and Science Foundation Ireland under grant 07/CE/I1147. The authors also wish to thank the staff at Poolbeg and Malahide Marinas for permission, assistance and access to the deployment location and facilities. Some students have assisted in this work and deserve thanks, Matthew Meagher for producing drawings and Camilla Nardi Pinto for testing and calibration.

4. Conclusions

References

The design, building and operation of a low-cost, multi-wavelength optical system for water quality monitoring has been illustrated and described. The OCS is capable of measuring the change in optical signal along two different optical paths (transmitted and sidescattered). The performance of the system has been tested in the laboratory with food dye and turbidity standard solutions of varying concentrations. These results show strong linear correlations between the optical signal response and turbidity concentration, as well as colour change from the food dyes. The system was deployed in the field at different locations and the recorded data correlated well with commercial water quality sensors. The results from the field trials highlight that the system has the ability to detect both sudden and significant changes in water opacity arising from environmental events. The work presented in this paper has outlined the first generation of the sensor development and results. Further work and potential adaptations are required to produce a sensor that can be easily and quickly calibrated in the field; along with more sophisticated data analysis, this will enable the sensor to produce the most desirable results from the data collected. In order to facilitate this, additional longer term deployments in test locations are planned. It is now possible to gather this type of data over a large spatial area, in real-time, at a reasonable cost; this would make these kinds of systems particularly attractive to lower income countries struggling with the effects of climate change and water related challenges. The intention is to commercialise the OCS as a real-time pollution monitoring system. Acknowledgements

[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11]

[12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22]

The authors acknowledge the funding of this research through the Beaufort Marine Research Award, carried out under the Sea Change Strategy and the Strategy for Science Technology and Innovation (2006–2013), with the support of the Marine Institute, funded under the Marine Research Sub-Programme of the National Development Plan 2007–2013 under grant BEAU/SENS/10.

[23] [24] [25]

European Water Framework Directive, Directive 2000/60/EC. European Bathing Water Directive, Directive 2006/7/EC. European Birds Directive, Directive 2009/147/EC. European Habitats Directive, Directive 1992/43/EC. D. Diamond, K.T. Lau, S. Brady, J. Cleary, Talanta 75 (3) (2008) 606–612. http: //dx.doi.org/10.1016/j.talanta.2007.11.022. R. Greenwood, J. Webster, F. Regan, Royal Society of Chemistry Report 2007. J. Grath, R. Ward, A. Scheidleder, P. Quevauviller, J. Environ. Monit. 9 (11) (2007) 1162–1175. http://dx.doi.org/10.1039/B710665N. C. Albaladejo, P. Sanchez, A. Iborra, F. Soto, J.A. Lopez, R. Torres, Sensors 10 (7) (2010) 6948–6968. http://dx.doi.org/10.3390/s100706948. P.W. Rundel, E.A. Graham, M.F. Allen, J.C. Fisher, T.C. Harmon, New Phytol. 182 (3) (2009) 589–607. http://dx.doi.org/10.1111/j.1469-8137.2009.02811.x. S.M. Glenn, T.D. Dickey, B. Parker, W. Boicourt, Oceanography 13 (1) (2000) 24–34. http://dx.doi.org/10.5670/oceanog.2000.50. T.S. Moore, K.M. Mullaugh, R.R. Holyoke, A.S. Madison, M. Yucel, G.W. Luther III, Annu. Rev. Mar. Sci. 1 (2009) 91–115. http://dx.doi.org/10.1146/annurev. marine.010908.163817. W. Bourgeois, A.C. Romain, J. Nicolas, R.M. Stuetz, J. Environ. Monit. 5 (6) (2003) 852–860. http://dx.doi.org/10.1039/B307905H. T. Leeuw, E.S. Boss, D.L. Wright, Sensors 13 (6) (2013) 7872–7883. http://dx. doi.org/10.3390/s130607872. B.H. Tangena, et. al., Water Contamination Emergencies (2011), pp. 13–31 . 10.1039/9781849733199-00013 A. Radu, S. Anastasova, C. Fay, D. Diamond, J. Bobacka, A. Lewenstam, Proceedings of the IEEE Sensors Conference, 2010, pp. 1487–1490. 10.1109/ICSENS.2010.5690357 A. Whelan, F. Regan., J. Environ. Monit.8 (9) (2006) 880–886. http://dx.doi.org/ 10.1039/B603289C. L.D. Chambers, et al., Surf. Coat. Technol. 201 (6) (2006) 3642–3652. http://dx. doi.org/10.1016/j.surfcoat.2006.08.129. W.F. Henley, M.A. Patterson, R.J. Neves, A. Dennis Lemly, Rev. Fish. Sci. 8 (2) (2000) 125–139. N.K. Goosen, J. Kromkamp, J. Peene, P. van Rijswijk, P. van Breugel, J. Mar. Sys. 22 (2–3) (1999) 151–171. A. Tremblay, L. Varfalvy, C. Roehm, M. Garneau (Eds.), Environmental Science and Engineering Series, Springer, New York. Y.A. Pachepsky, D.R. Shelton, Crit. Rev. Environ. Sci. Technol. 41 (12) (2011) 1067–1110. N.W. Bergmann, M. Wallace, E. Calia, Proceedings of the Sixth International Conference on Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP), 2010, pp. 19–24 . 10.1109/ISSNIP.2010.5706802 J. Trevathan, et al., Sensors 12 (7) (2012) 9711–9748. http://dx.doi.org/10.3390/ s120709711. A.A. Gilerson, et al., Opt. Express 18 (23) (2010) 24109–24125. http://dx.doi. org/10.1364/OE.18.024109. F.E. Hoge, R.N. Swift, Appl. Opt. 25 (15) (1986) 2571–2583. http://dx.doi.org/ 10.1364/AO.25.003677.