Environ Monit Assess (2022) 194:122 https://doi.org/10.1007/s10661-021-09738-z An innovative autonomous robotic system for on‑site detection of heavy metal pollution plumes in surface water Elisabetta De Vito‑Francesco · Alessandro Farinelli · Qiuyue Yang · Bhawna Nagar · Ruslan Álvarez · Arben Merkoçi · Thorsten Knutz · Alexander Haider · Wolfgang Stach · Falko Ziegenbalg · Roza Allabashi Received: 22 April 2021 / Accepted: 23 December 2021 © The Author(s) 2022 Abstract Smart monitoring has been studied and using carbon-based screen-printed electrodes (SPEs). developed in recent years to create faster, cheaper, The focus of this work was to validate the ability and more user-friendly on-site methods. The present of the integrated system to perform on-site detec- study describes an innovative technology for investi- tion of heavy metal pollution plumes in river catch- gative monitoring of heavy metal pollution (Cu and ments. This scenario was simulated in laboratory Pb) in surface water. It is composed of an autono- experiments. The main performance characteristics mous surface vehicle capable of semiautonomous of the system, which was evaluated based on ISO driving and equipped with a microfluidic device for 15839 were measurement bias (Pb 75%, Cu 65%), detection of heavy metals. Detection is based on the reproducibility (in terms of relative standard devia- method of square wave anodic stripping voltammetry tion: Pb 11–18%, Cu 6–10%) and the limit of detec- tion (4 µg/L for Pb and 7 µg/L for Cu). The lowest Supplementary Information The online version detectable change (LDC), which is an important contains supplementary material available at https://​doi.​ performance characteristic for this application, was org/​10.​1007/​s10661-​021-​09738-z. E. De Vito‑Francesco (*) · W. Stach · F. Ziegenbalg · B. Nagar R. Allabashi Laboratory of Physical and Analytical Electrochemistry Department of Water, Atmosphere, and Environment, (LEPA), Ecole Polytechnique Fédérale de Lausanne Institute of Sanitary Engineering and Water Pollution (EPFL) Valais Wallis, Rue de l’Industrie 17, 440, Control, University of Natural Resources and Life Sciences 1951 Sion, Switzerland Vienna (BOKU), Muthgasse 18, 1190 Vienna, Austria e-mail: elisabetta.de-vito-francesco@boku.ac.at A. Merkoçi Catalan Institution for Research and Advanced Studies A. Farinelli (ICREA), Passeig Lluís Companys 23, 08010 Barcelona, Department of Computer Science, University of Verona, Spain Ca Vignal 2, 37134 Verona, Italy T. Knutz · A. Haider Q. Yang · B. Nagar · R. Álvarez · A. Merkoçi Go Systemelektronik GmbH, Falunerweg 1, 24109 Kiel, Catalan Institute of Nanoscience and Nanotechnology Germany (ICN2), UAB Campus, 08193 Bellaterra (Barcelona), Spain Q. Yang Materials Science, Department of Chemistry, Universitat Autònoma de Barcelona, Plaça Cívica, 08193 Bellaterra (Barcelona), Spain Vol.: (0123456789) 13 122 Page 2 of 19 Environ Monit Assess (2022) 194:122 estimated to be 4–5 µg/L for Pb and 6–7 µg/L for Cu. Consistent and smart monitoring contribute to The life span of an SPE averaged 39 measurements the preservation and protection of the aquatic envi- per day, which is considered sufficient for intended ronment and avoid negative effects on human health monitoring campaigns. This work demonstrated the (EEA et al., 2018; EU, 2000, 2013; Melo et al., 2019; suitability of the integrated system for on-site detec- Odobašić et al., 2019). The traditional monitoring tion of Pb and Cu emissions from large and medium methodology consists of discrete or cumulative, man- urban areas discharging into small water bodies. ual or automatic sampling, followed by laboratory analysis (Berho et al., 2009; EEA, 1996; Steccanella Keywords Heavy metal pollution · Surface et al., 2019). Within this monitoring process, sam- water monitoring · Square wave anodic stripping pling and laboratory analysis must be included. The voltammetry · Screen-printed electrode · Autonomous most commonly used analytical methods for heavy surface vehicle metal analysis are inductively coupled plasma mass spectrometry (ICP-MS), inductively coupled plasma atomic emission spectrometry (ICP-AES), atomic Introduction absorption spectroscopy (AAS), and atomic fluo- rescence spectroscopy (AFS) (Gumpu et al., 2015; Water is of fundamental importance for the environ- Koller & Saleh, 2018; Verma & Singh, 2005). These ment and life, and it determines the health of humans conventional techniques achieve high accuracy and and animals (Silva Junior et al., 2016; Tuna et al., sensitivity in measurements, but require a sampling 2013). Both natural processes, such as weathering of procedure and a well-trained analyst. Furthermore, soil, and anthropogenic activities, such as rapid indus- they are time-consuming and expensive (Berho et al., trialization, urbanization and agricultural activities, 2009; Butterfield, 2009; Gumpu et al., 2015; Verma deeply influence the contamination of water resources & Singh, 2005). with a variety and a large number of contaminants Faster, less expensive and more user-friendly on- (Law et al., 2018; Li et al., 2015; Mohammed et al., site methods are increasingly playing a crucial role in 2016). In particular, there is great concern regard- the field of quality monitoring of surface water bodies ing the environmental fate of heavy metals (HMs) (Berho et al., 2009). In this regard, electrochemical and their impacts on human health, which are related techniques are very promising due to their sensitivity, to bioaccumulation, non-biodegradability, persis- low cost, accuracy, simplicity and on-site applicabil- tence and toxicity (Gautam et al., 2014; Güell et al., ity (Berho et al., 2009; Lu et al., 2018; Waheed et al., 2008; Gumpu et al., 2015; Jang et al., 2011; Law 2018). Voltammetry, particularly square wave anodic et al., 2018; Li et al., 2015; Masindi & Muedi, 2018; stripping voltammetry (SWASV), is the technique Odobašić et al., 2019; Waheed et al., 2018). The abil- usually applied to electrochemical measurements. Its ity of metals to pass through cell membranes causes virtues, relative to other techniques, include higher several harm for different physiological processes of sensitivity, selectivity, shorter detection time and cells and therefore poses a threat to human and ani- lower costs (de la Escosura-Muñiz et al., 2008; Güell mal health (Gumpu et al., 2015; Jang et al., 2011). et al., 2008; Lu et al., 2018; Merkoçi et al., 2000; The majority of heavy metal releases into water bod- Waheed et al., 2018). Printing technique, as massive ies in Europe are related to urban pollution coming production with low cost, has brought the fabrication from urban wastewater treatment plants (UWWTPs, of micro- and nano- sensors out of cleaning rooms 45.5%), from misconnections and cross-connections, and has been harnessed widely in sensing applications untreated combined sewer overflows or road run- especially in stack structure (Wiklund et al., 2021). off (EEA, 2018). Furthermore, the majority of the However, due to the difficult manipulation of dif- sources of heavy metals to these UWWTP come ferent inks to be printed on the stack structure, fully from industries (78.4%) (EEA, 2018). Heavy met- printed sensors and sensing systems are still an open als are considered indicators for the abovementioned challenge (Khan et al., 2020). To avoid the limitations emission sources, and their monitoring is of crucial associated with bulky electrodes usually employed importance in identifying the sources and pathways for voltammetry, the use of fully screen-printed elec- of potential pollutants in the aquatic environment. trodes has recently been applied for on-site analysis Vol:. (1234567890) 13 Environ Monit Assess (2022) 194:122 Page 3 of 19 122 (Berho et al., 2009; Güell et al., 2008; Waheed et al., integrated system: the autonomous surface vehicle 2018). An additional positive feature of anodic strip- and the HM detection device. Second, a comprehen- ping voltammetry for environmental monitoring is sive validation involving laboratory and field experi- its ability to detect the soluble ionic form of metals ments is performed to establish the performance (Barón-Jaimez et al., 2013; Zinoubi et al., 2017), characteristics of the system. The validation was car- which is the fraction responsible for metal bioavail- ried out based on the guidelines available in the ISO ability in the aquatic environment (Rensing & Maier, 15839 standard (International Standard, 2003, 2019). 2003); this quantity is targeted in the field of water The validation process included the evaluation of policy for regulation of priority substances (EU, the integrated system during the simulation of real 2000, 2013) in the framework of the European Water scenarios in the laboratory facility, followed by on- Frame Directive (EU, 2000). site campaigns to test the simultaneous operation of To overcome issues arising from conventional autonomous navigation and measurement. Finally, the monitoring methods, including the high costs and validation results are further discussed in terms of the lack of information between sampling campaigns, an fit-for-purpose status of the integrated system, which alternative monitoring method involves, in addition depends on its ability to detect the impact of certain to fixed monitoring stations, the use of autonomous pollution sources on receiving waters. surface vehicles (Steccanella et al., 2019). These sys- tems consist of mobile, autonomous robotic systems, which can be programmed to perform activities inde- Material and methods pendently; for instance, they can engage in sampling or real-time/on-site analysis and measurements of the Autonomous surface vehicle matrix to be monitored (Melo et al., 2019). The need for new monitoring concepts for detec- In the aforementioned project INTCATCH 2020, a tion of pollutants in surface bodies is included in the commercial model of the autonomous surface vehi- vision of the Horizon 2020 European-funded project cle (ASV; Fig. 2b) Lutra Prop®, was used. This is a INTCATCH. The project emphasizes the need to monohull boat mounted with submerged propellers create smarter monitoring concepts using innovative and measuring approximately 1 m long and 0.5 m technologies and holistic concepts. The present paper wide, and it was developed by Platypus® (Steccanella describes one of the monitoring systems developed et al., 2019). The platform has been engineered and within that framework, which consists of an autono- optimized within the INTCATCH project to include mous surface vehicle equipped with a microfluidic several additional sensors and advanced autonomy electrochemical heavy metal detection device, named capabilities. The main components of the architecture integrated system. The main aim of the present study of the electronic system are an e-board, a measure- is to demonstrate that this integrated system is fit-for- ment and sensor manager (Bluebox®), a smartphone purpose in monitoring wastewater point emissions in and an external control unit device (tablet). A sche- surface waters, which means it is capable of detecting matic representation of the architecture of the boat is heavy metal pollution plumes in surface waters com- shown in Fig. 1. The architecture of the ASV allows ing from wastewater spill-outs. Lead and copper were the end user to control the navigation feature settings selected as target metals for system validation. Lead remotely, start on-site measurements and start the is included in the list of priority substances from the sampling of the monitored matrix. The boat is pow- Water Framework Directive (WFD) (EU, 2000, 2013), ered by one lithium polymer battery (4 S, 16 Ah, 10 and copper is a known representative of the pollut- C, 16–14 V). ants released into water bodies from mining activities Incoming and outcoming control signals for the (EEA, 2018), industry (EEA, 2018) and stormwater autonomous surface vehicle are managed by the runoff (Gromaire et al., 2001; Müller et al., 2019; e-board. It is the core of the autonomous surface Winters & Graunke, 2014; Winters et al., 2015). The vehicle and is composed of a Platypus circuit board present study first describes the components of the and an Arduino Due® board. It connects to and Vol.: (0123456789) 13 122 Page 4 of 19 Environ Monit Assess (2022) 194:122 provides power for all the other components. Blue- features (customizable design of waypoints, path- Box® is a measurement data manager that manages ways and velocity levels) and the measuring system the interpretation of the measured electrical signals device. With this option, the end user can initiate the and is directly connected to the implemented meas- HM measurement by choosing between “single” and uring device. Furthermore, it uploads and stores the “continuous measurement” modes. The first consists measurement data in the Cloud Database and locally of the performance of a single measurement, and the on the Bluebox® system. The smartphone is used second comprises consecutive and continuous rep- as the main computation component of the ASV, etition of single measurements. Another possibility controlling its autonomous functions and receiv- offered by the GUI is the autonomous sampling of the ing navigation and measurement commands via a monitored matrix (Fig. 2b). This consists of the inde- WiFi connection with an external control unit (tab- pendent filling of four sampling containers integrated let). The latter is equipped with a graphical unit into the autonomous surface vehicle (0.5 L volume interface (GUI). Through the GUI, the end-user can each). These water samples are needed first for heavy load live maps of the selected catchment and there- metal reference analysis and second for analysis of fore to visualize the position of the ASV and user. It additional parameters to complete the water quality is possible to send commands to control navigation information. Fig. 1  Schematic represen- tation of the architecture of the autonomous surface vehicle’s control system, including the options for manual remote control (RC) Vol:. (1234567890) 13 Environ Monit Assess (2022) 194:122 Page 5 of 19 122 Heavy metal detection device the potentiostat (Fig. 2d). In comparison with a three-electrode system, in a two-electrode sys- The autonomous surface vehicle is equipped with tem the reference and the counter electrodes are a microfluidic detection device to provide for HM merged into one electrode. In this case, it is more on-site detection (Fig. 2a). This system consists difficult for the reference electrode to provide a of a peristaltic pump, an automatic mixing sys- stable reference potential and simultaneously col- tem, a bubble trap, a flow cell and a potentiostat lect electrons as counter electrodes. Therefore, (Fig. 2c). A carbon paste screen-printed electrode the three-electrode electrochemical system was (SPE), including a miniaturized three-electrode selected as suitable for the integrated system. The electrochemical system as the main sensing ele- microfluidic detection device is compatible with ment, is fixed in the flow cell and connected with the electrochemical method of square wave anodic Fig. 2  a Implementation of filtration system and HM microfluidic detection device on the autonomous surface vehicle; b autono- mous surface vehicle, equipped with HM micro- fluidic detection device, filtration system, and sampler, deployed in a field experiment; c schematic representation of compo- nents of the HM micro- fluidic detection device; d schematic representation of one SPE; e example of the voltammogram Vol.: (0123456789) 13 122 Page 6 of 19 Environ Monit Assess (2022) 194:122 stripping voltammetry (SWASV), generally cou- Measurement and quality assurance pled with a three-electrode electrochemical system. First, through the application of a specific negative The raw data of a single measurement performed potential on the working electrode, the HM ions with the HM device consisted of the measured elec- are concentrated, accumulated and reduced with trical current related to the corresponding potential the flow passing through the flow cell and over the (Fig. 2e). The potentials at which electrical current SPE. Subsequently, the potential is released to 0 by peaks occurred and the integrated areas below those square waves used to strip back the HM ions. The peaks were automatically computed by the software raw data, or voltammogram, consist of a plot of the used. Before each experimental campaign, a fresh measured electrical current obtained from HM ions screen-printed sensor was employed and calibrated. stripping (mA) as a function of the potential (V). Representative voltammograms are shown in Fig. 10 The various peak potentials are assigned to differ- in the Supporting information. ent heavy metal species, as in the example shown The calibration curves were generated from meas- in Fig. 2e), where the first peak corresponds to Pb urements of mixed solutions of both metals, Pb and and the second to Cu. The intensity of the current Cu, performed with the described integrated system signal is proportional to the concentration of the (microfluidic detection device implemented on the analytes in the sample. In our study, the integrated ASV). The calibration solutions were first prepared area below the electrical current peak (µAV) was within a concentration range of 10–100 µg/L, keeping used as the output signal. The HM microfluidic the 1:1 ratio for both metals. For a better representa- detection device does not perform sample stir- tion of the concentration ratio of these metals in real ring or any pretreatment of the samples other than surface waters, some experiments were performed filtration (Fig. 2a) and mixing with the electrolyte with mixed solutions containing a 1:4 ratio (Pb:Cu), needed for analysis. The sample is filtered through leading to a working range of 40–400 µg/L for Cu. two in-sequence nylon net (Millipore) filters with Two kinds of linear regression functions, linear 180 µm and 80-µm net pores and of 47 mm diam- forced and non-forced through zero and a non-linear eter. The presence of an electrolyte (HCl) increases regression function, were considered regarding their the release of metal ions present in the water sample influence on the measuring accuracy. from bound forms and helps with the transportation To perform quality assurance of the measurements, of ions during the measurement. The ratio of the mixed control samples (CS) of Pb and Cu were prepared sample to electrolyte was set to 7:1 to obtain a final with a concentration at the middle of the defined work- concentration of 0.05-M HCl. ing range. The control samples were prepared in the Since the integrated system is still a proof of con- lab from stock solutions (1000 mg/L Pb, as Pb(NO3)2 cept as a fully automatic tool to simultaneously detect in ­H2O, Titrisol®, 1.09969.0001 and 1000 mg/L Cu, as multi-HMs, lead and copper were chosen as model ­CuCl2 in ­H2O, Titrisol®, 1.09987.0001, VWR). Both HMs and target heavy metals for the system valida- calibration solutions and control samples were always tion. During the experiments performed to validate measured with the microfluidic detection device imple- the detection of all representative heavy metal pol- mented in the integrated system. lutants, such as zinc (Zn), cadmium (Cd), lead (Pb) and copper (Cu), it turned out that Cd presents a high Validation procedure limit of detection (LOD) due to the mutual interfer- ence with other heavy metals and the potential for The methodology used to validate the integrated sys- Zn (~ −1.2 V) does not fit in our SWASV sensing tem was based on the ISO 15839 protocol (reviewed potential window/range (from −1.0 to 0 V). Conse- in 2019) “Water quality – online sensor/analysing quently, we focused on Pb and Cu for the purpose of equipment for water – specifications and performance this study. Any further details concerning the set-up tests” (International Standard, 2003, 2019). The named and the sensing performance of the HM microfluidic standard was reviewed and confirmed in 2019; there- detection device can be found in the publication from fore, the current version can be considered to be used. Yang et al. (2021). Vol:. (1234567890) 13 Environ Monit Assess (2022) 194:122 Page 7 of 19 122 This standard can be applied to most sensors/ the laboratory experiments were used to calculate the analysis equipment, but it is recognized that for some aforementioned performance characteristics. equipment, certain performance tests cannot be car- ried out (International Standard, 2003, 2019). There- Laboratory experiments fore, the validation procedure for the equipment under study (integrated system) was focused on the Laboratory experiments were designed and performed assessment of performance characteristics, which to validate the integrated system for use with the case were especially useful to describe the suitability of of a potential spill-out of polluted wastewater into a sur- the system for the intended purpose. As the intended face water body. This situation was simulated in a labo- purpose is the detection of heavy metal pollution ratory set-up and is shown in Fig. 3. A 500-L water tank plumes in surface waters, the most important per- was filled with water collected from the Danube River formance characteristic is considered to be the low- or from a groundwater well, which represented bank- est detectable change (LDC), which is defined as the filtrated Danube River water. The introduction of pre- “smallest significantly measurable difference between pared solutions of lead (1000-mg/L Pb, as stock solu- two measurements”. This characteristic gives infor- tion) and copper (1000-mg/L Cu, as stock solution) into mation about the ability of the measuring system to the real water matrix was used to simulate point-source detect spatial concentration gradients in a defined contamination. The matrix was spiked at defined time area around the pollution source. Additionally, the intervals of approximately 60 min (Fig. 3). The desired limit of detection (LOD) and limit of quantification theoretical concentrations for each spike level, as in the (LOQ) were evaluated according to the test procedure example shown in Fig. 6, were 25, 50, 75 and 100 µg/L described in ISO 15839. Day-to-day repeatability for Pb and 100, 200, 300 and 400 µg/L for Cu. The was chosen to address measurement precision. The number of spiking levels and the desired concentrations evaluation of systematic error (bias) was performed varied slightly between experiments. To avoid the sedi- mainly through recovery experiments, which involved mentation of suspended particles, a stirrer was situated the extent of recovery of a known amount of Pb and within the tank and operated at 1000 rpm. The inte- Cu added to a previously analysed sample (Interna- grated system was located in the tank and programmed tional Standard, 2012). A comparison of measured to perform measurements in continuous mode during values with two selected reference methods was also the whole simulation. For the reference analysis of Pb performed for this purpose and discussed in the sec- and Cu in the lab, duplicates of samples were collected tion “Reference analysis”. The data acquired from 30 min after each contamination spike. Fig. 3  Schematic repre- sentation of the laboratory experimental procedure Vol.: (0123456789) 13 122 Page 8 of 19 Environ Monit Assess (2022) 194:122 Reference analysis Field experiments Photometry was identified as the proper reference Field experiments consisted of deploying the inte- method for comparison and calculation of recovery grated system at a selected location and testing its because of its ability to analyse free ions in water, abilities in a real environment. The main goal of the as well as the SWASV method, used in the inte- performed field experiments was to test the simul- grated system. Therefore, similar results are expected taneous functioning of the main features of the between the photometric method and the SWASV integrated system while the ASV was set to autono- one. Furthermore, the photometry method was con- mous navigation; these involved continuous meas- sidered for its portability, and therefore a potential urements with the HM microfluidic detection device on-site reference method. The kits “copper trace and automatic filling of the vessels with the auto- cuvette test” and “lead trace cuvette test” (Hach matic sampler device. Lange GmbH¸ LCK529 and LCK306, respectively) The selected site was located within the City of were used to perform the reference analyses without Vienna, at the New Danube River, in the vicinity of any pretreatment of the water samples. The bathocu- an oil refinery (48° 11′ 09.8″ N 16° 28′ 37.1″ E). proine disulphonic acid method and the PAR method Prior to deployment, the calibration of the screen- (4-(2-pyridylazo)-resorcinol) were applied for Cu printed sensor was performed in the laboratory. and Pb, respectively. The working ranges for these The set-up of the integrated system was made pos- kits are 0.01–1.0 mg/L and 0.1–2.0 mg/L for Cu and sible with the GUI. The speed was set at “low speed”, Pb, respectively (HACH, 2019, 2020). which corresponded to ~ 1 m/s. The path (waypoints) was Even though inductively coupled plasma mass designed to cover an area of ~ 100 × 15 m of the water spectrometry (ICP-MS) is known to analyse the body. The HM detection device was set to “continuous total amount of heavy metals in water, therefore measuring mode”. The measurement time was established less feasible than the photometric method, it was to be approximately 2 h for each deployment. Each single additionally selected as a reference method for two HM measurement lasted ~ 7 min, therefore allowing ~ 17 reasons. First, in the integrated system, the sample on-site measurements. The values measured were stored was premixed with acid prior to SWASV measure- in Bluegate (the cloud of the Bluebox® data manager). ment so that a certain amount of ions were released In addition, the collection of four water samples from their bound state into the free state (acidifying for each campaign was possible with the integrated process (Güell et al., 2008; Palchetti et al., 2005)). sampling device, as described in the “Autonomous In this case, the content of metal ions measured surface vehicle”. was expected to be higher than that measured by the photometric method. Therefore, depending on the release rate, the quantity measured by the inte- Results and discussion grated system (SWASV) could be better compared to that of ICP-MS. Second, the chosen photometric Calibration function method fails to measure concentrations smaller than 0.01 and 0.1 mg/L for Cu and Pb, respectively. In The analysis of Pb and Cu using screen-printed elec- these cases, ICP-MS was the only possible analytical trodes (SPE) coupled with SWASV was based on a method available as a reference analysis. The instru- calibration function generated prior to sample analysis. ment Elan DRC-e (PerkinElmer Inc.) was used to The values of the measured peak areas (µAV, two rep- perform this analysis by applying the standard meth- resentative voltammograms are included in Fig. S1) for ods DIN EN ISO 17294–1 (E36)-2003 for the appli- the selected calibration solutions were plotted in rela- cation of the ICP-MS and DIN EN ISO 17294–2 (E tion to the known concentration values of these solu- 29)-2004 for the selected metals. The limits of quan- tions. The calibration function was then obtained by tification (LOQs) of the method used were 0.5 µg/L fitting the plotted points with a regression trend. As the and 1.0 µg/L for Pb and Cu, respectively. calibration function strongly influenced the accuracy Vol:. (1234567890) 13 Environ Monit Assess (2022) 194:122 Page 9 of 19 122 of the final measurement, its role was first evaluated. In the case of SWASV, the limited lifetimes of SPEs made this crucial, as the calibration function related to each specific SPE had to be changed daily for new experiments and/or measuring campaigns. Three regression models were considered for the calibration function: linear, linear forced through zero and non-linear (polynomial of second order). Using a large number of calibration functions performed (40 in total), it was observed that all three calibration mod- els exhibited good fits, with R2 > 0.90 for most cases. Even though the polynomial model showed better val- ues (R2 > 0.98), fits with both linear calibration mod- els were considered satisfactory for this application. Therefore, the linear regression model was selected for further computations. The following labels were used throughout the study: “cal1” indicated the cali- bration function obtained with linear regression, and “cal2” indicated the calibration function obtained with linear regression forced through zero. It was further observed, as expected, that the higher intercepts of calibration functions (cal1) led to inaccurate values at low concentrations of heavy metals in samples. The authors decided to use the forced-through-zero calibra- tion function (cal2) for further calculations; although Fig. 4  Recovery of the control sample (CS) for Pb and Cu and this is not the way to improve measurement accuracy, for ratios of 1:1 and 1:4, calculated with both linear calibra- it is a reasonable compromise for the analytical method tions cal1 (linear regression) and cal2 (linear regression forced used and the purpose of this study. through 0). The mean values of the recoveries are represented In Fig. 4, the recovery (%) of the control samples by the cross symbols for Pb and Cu, computed with both cal1 and cal2, is shown. The recovery for Pb was between 70 and 73% sampling application in combination with the purpose (median values, ratio 1:1) and 48 and 54% (median of this study and the detection of wastewater source- values, ratio 1:4), and for Cu it was between 58 and point emissions in surface water. In such cases, a pol- 70% (median values, ratio 1:1) and 38 and 42% lution plume is expected, and the most important per- (median values, ratio 1:4). The selection of the cali- formance characteristic for detection is considered to bration function did not have a significant influence be the lowest detectable change, rather than measure- on the accuracy of Pb and Cu measurements. This ment accuracy. can be explained by the fact that the concentrations of control samples were selected to be in the middle Sensor lifetime part of the calibration range, where the influence of the changed intercept is expected to be minimal. The main evaluation of the analytical performance of The authors are aware that both characteristics, the SPEs under laboratory conditions was already per- linearity of the calibration function and the accuracy formed by the system developers and is reported in in terms of control sample recovery, are rather poor Yang et al. (2021). In the context of this work, only compared to those of standardized laboratory meth- the lifetime of the SPE involved in the integrated ods. However, they can be considered acceptable system and in use during the measuring campaigns for the aforementioned compromise between on-site was observed and documented. This is considered Vol.: (0123456789) 13 122 Page 10 of 19 Environ Monit Assess (2022) 194:122 an important factor for field monitoring campaigns Validation through laboratory experiments because of the limited number of measurements allowed per SPE and the necessity to use a fresh SPE The experiment reported in Fig. 6 consisted of the for each measuring campaign (day). The results are simulation of four contamination levels in the range shown in Fig. 5, where each bar shows the number of 0–100 µg/L for Pb and 0–400 µg/L for Cu and use of measurements performed with one SPE during an of bank-filtrated Danube River water as the matrix, entire laboratory or field experiment, including the and it is considered a representative example of seven preliminary calibration. The maximum number of similar laboratory experiments performed. measurements performed with 1 SPE was 55, and the Figure 6 shows the Cu and Pb concentrations minimum number was 16. It must be outlined that measured by the integrated system (IS) and com- these low values (16, 19, 20 measurements) were puted with cal2, the mean concentration values for not related to poor quality or damaged sensors but each spike level and results of reference analyses. simply to other technical issues arising during the Table 1 summarizes the results obtained during the experiments (e.g. malfunctioning of the integrated indoor experiments. The difference between the mean system during the experiment), which required a concentration values of consecutive levels should be change of sensor. The calculated average value of constant, as the water matrix was spiked with the 39 measurements per sensor, which included an ini- same amount of contaminants at each level. For Cu, tial quality assurance check (triple measurement), generation of the calibration function needed (tripli- cate measurement of five calibration solutions) and finally, effective monitoring of the water body, was considered sufficient for the intended monitoring campaigns. Fig. 5  Number of measurements performed with each SPE and average value and average value represented by the hori- Fig. 6  Laboratory experiment results, including lead (orange) zontal line and copper (blue) in a ratio of 1:4 and reference analyses Vol:. (1234567890) 13 Environ Monit Assess (2022) 194:122 Page 11 of 19 122 Table 1  Indoor Theor. AV AV AV Recovery Recovery Recovery experiment: average (AV) concentration conc. conc. conc. (theor. (ICP- (LCK) concentrations (conc.) conc.) MS) detected by the integrated IS ICP-MS LCK system (IS) and by the Unit µg/L µg/L µg/L µg/L % % % two reference methods (ICP-MS and LCK) and Pb recoveries relative to the Level 1 0.0 (4.7) < 0.5 (15.5) - - - theoretical (theor.) value Level 2 25.0 16.4 15.9 30.0 66 103 55 and the measured reference Level 3 50.0 22.2 32.6 32.0 44 68 70 values (ICP-MS and LCK). Values in brackets are Level 4 75.0 30.8 48.9 50.0 41 63 62 values < LOQ Level 5 100.0 35.6 64.2 66.5 36 56 54 Cu Level 1 0.0 (2.9) 10.5 (9.0) - - - Level 2 100.0 40.9 62.5 63.5 41 65 64 Level 3 200.0 77.4 120.0 126.5 39 65 61 Level 4 300.0 108.4 173.0 185.5 36 63 58 Level 5 400.0 123.2 226.5 250.0 31 54 49 the concentration differences of consecutive levels fulvic substances, serves as the main source for showed a rather constant behaviour for the first three organic ligands able to build stable complexes with levels (~ 35 µg/L differences, instead of the theoreti- most metal ions (Boggs et al., 1985; Mostofa et al., cal difference of 25 µg/L) and some discrepancy for 2013a; Mostofa, 2013). These compounds exhibit the last level (only ~ 14 µg/L increase). Severe vari- a strong interaction with ion metals, forming com- ability was observed for Pb, in which this difference plexes with covalent bonds (Boggs et al., 1985; decreased for each spike level. Nevertheless, the Mostofa et al., 2013b; Pandey et al., 2000). These integrated system was able to detect the change in substances are always present in surface waters concentration for both metals at each contamination and, according to Mostofa et al. (2013b), comprise level. between 20 and 85% of the DOM in rivers. Humic As seen in Table 1 and in Fig. 6, significant dis- substances are also the main component in DOM crepancies were observed between measured values of groundwater environments (Pisarek & Głowacki, determined with the integrated system and refer- 2015; Steinberg et al., 2003). These substances also ence methods. The calculated recoveries for Pb were play an essential role in the complexation of metal 56–103% and 54–70% and those for Cu were 54–65% ions in water, thereby decreasing their availability and 49–64%, relative to ICP-MS and LCK results, as free ions (Pisarek & Głowacki, 2015; Steinberg respectively. The reference methods showed similar et al., 2003). Considering that the ICP-MS method values to each other while showing a discrepancy provides the total ion concentration (free and com- with the SWASV method applied to the integrated plexed form), the discrepancy between the ICP-MS system. One possible explanation for this could be the and integrated system results may be explained by existence of various metal species in the water matrix the presence of metal complexes of humic or fulvic and the abilities of the analytical methods to address substances in the river water used for experiments. them differently. The gap between the concentrations measured with The employed SWASV method detected the the integrated system and the ICP-MS reference free metal ions soluble in water (Barón-Jaimez method indicates that a certain part of Pb and Cu in et al., 2013; Zinoubi et al., 2017), but not the frac- the river water samples could not be released from tion bonded in complexes and other stable chemi- the organic complexes before measurement, despite cal compounds. Most metals have the ability to the pre-acidification with HCl. One possible reason form complexes with organic substances or other for this may be the short contact time of HMs and compounds present in the water matrix. Dissolved HCl in the described micro fluidic detection device. organic matter (DOM), in particular humic and Vol.: (0123456789) 13 122 Page 12 of 19 Environ Monit Assess (2022) 194:122 The photometric method was selected as a refer- development of the integrated system, a deposition ence method based on its ability to detect free metal time of 200 s was chosen in view of the need for a ions in water without any sample pretreatment; hence, short measurement time. It can be assumed that the it is expected to give results similar to those from the limited time available for ion deposition might be SWASV method. Nevertheless, a significant differ- responsible for observed discrepancies from the pho- ence in detected concentrations was also observed tometric method. between these two methods, with significantly lower The good comparability between ICP-MS and results for SWASV. This could be attributed to two photometry (LCK) measurements is a strong indica- factors. First, the real sample matrix may influence tion of the limited formation of complexes between the sensing capability through a possible deposition metal ions and humic substances. Therefore, the lim- of DOM onto the surface of the working electrode ited deposition capability of SWASV for metal-free during the deposition step. This would enhance the ions may explain the bias between the integrated sys- background current and block active sites, lead- tem and both reference methods. An additional con- ing to a reduced signal for the heavy metal (Borrill firmation was provided by the higher bias observed et al., 2019; Lam et al., 1997). From our observation, for higher (within the ranges used) ion concentrations the measurement baseline for Danube river samples for both metals and reference methods, as presented kept shifting higher with an increase in the number of in Fig. 7. measurements, which may be caused by the described This leads to the conclusion that the selected ana- effect of DOM. Second, the lower results for the lytical methods cannot provide the agreed reference SWASV method could be related to the capability of value needed for evaluation of bias in the electro- the screen-printed electrode to collect and detect the chemical SWASV method used in this application. free ions from water, which would, in turn, be related The authors decided to use the theoretical concentra- to the settings of the micro fluidic detection device tion value from the self-prepared solutions (control and mainly the deposition time interval selected. samples) for the evaluation of this characteristic. The integrated areas for Cu and Pb in a mixed stand- ard solution increased with longer deposition times in the range of 60 to 400 s. However, during the Validation through field experiments Three outdoor experiments were carried out to observe the behaviour and the performance of the integrated system under field conditions. The water quality of the selected catchment did not allow to observe any change in the concentrations of lead and copper dur- ing the measurement campaign. The total concentra- tions of Pb and Cu in all collected samples analysed by the ICP-MS method were below the limit of quantifi- cation. Nevertheless, during the outdoor experiments, it was possible to test the simultaneous operation of the autonomous features of the integrated system: the continuous measurement of the water matrix through the microfluidic device, the automatic collection of the water matrix through the sampler, and the autonomous navigation of the ASV. Several customized navigation paths were created with the use of the GUI, and the integrated system successfully responded to autono- mous navigation by following the path created (Fig. 8). The range of connectivity (between the integrated Fig. 7  Difference between values measured by reference methods and the integrated system (IS) for each contamination system and the GUI) was evaluated to be less than level a 100-m radius. However, the lack of connection Vol:. (1234567890) 13 Environ Monit Assess (2022) 194:122 Page 13 of 19 122 Fig. 8  Graphical user interface and example of the pathway designed on the loaded live map of the selected site during one of the field campaigns between the GUI and the autonomous surface vehi- filled independently and simultaneously after specific cle, once the 100-m limit of radius was reached, did commands were used with the GUI. not influence the autonomous behaviour of the inte- The study from Wang et al. (2017) present a simi- grated system: the autonomous surface vehicle con- lar autonomous surface vehicle (kayak) equipped with tinued navigating, and the microfluidic system device an on-site Zn detection, using liquid crystal polymer- kept performing the planned measurements. When the based electrodes coupled with SWASV. However, system was out of range from the GUI, only the func- this autonomous monitoring device needed on shore tionality of the latter was affected: the operator can- commands, for the ASV’s navigation and for the per- not change parameters for the autonomous behaviours formance of any electrochemical measurement, com- (i.e. change waypoints or path) and can control the mands which are both automized on the integrated boat only through a remote control (RC) unit. Future system of the present study (Wang et al., 2017). developments of the autonomous surface vehicle will include the use of 3G/4G technology to connect the GUI to the autonomous surface vehicle to avoid issues Table 2  Average values (AV) of the performance characteris- tics of the integrated system and number of experiments (n) with the connectivity range. Furthermore, the perfor- mance of the HM detection device was tested through- Performance characteristic Pb Cu out autonomous navigation. It was run in “continuous Unit AV n AV n measurement” mode and showed that all measure- ments were successfully completed, and the results Limit of detection µg/L 4 157 7 157 were saved in the BlueGate cloud and kept available Limit of quantification µg/L 14 157 22 157 for further computation. The sampling executed by the Lowest detectable change µg/L 4–5 77 6–7 81 sampling device, which was implemented on the inte- Recovery % 75 280 65 280 grated system, was also successful. The vessels were Reproducibility % 11–18 109 6–10 70 Vol.: (0123456789) 13 122 Page 14 of 19 Environ Monit Assess (2022) 194:122 Validation results reproducible performance (day-to-day repeatability of 5%), than the laser-pyrolized paper electrodes coupled The validation of the integrated system was partly with SWASV. This can be due to the different elec- based on the guidelines available in the ISO 15839 trode type and different voltammetry settings. Table S1 standard (International Standard, 2003, 2019). The shows the comparison between the performance char- performance characteristics of the system were com- acteristics of the different devices. puted by employing data collected during the labo- The authors are aware that the LOQ values of ratory experiments. As established in “Calibration the integrated system are not low enough to address function”, all selected data were calculated with cal2. the limits of environmental quality standards (EQS) The limit of detection (LOD), limit of quantification established by the Water Framework Directive (EU, (LOQ) and lowest detectable change (LDC) are con- 2000, 2013). While no specification can be found sidered as the most important characteristics for this there for Cu, the annual average EQS for lead and its application. The recovery was used to represent the compounds in inland surface waters is specified to be accuracy of the measurement. Finally, the measure- 0.02 µg/L, and the maximal allowable concentration is ment precision was described through day-to-day defined by 14 µg/L. Consequently, it can be concluded repeatability (reproducibility). The obtained results that the presented system is not suitable for monitor- are listed in Table 2. ing and assessment of the chemical status of surface Berho et al. (2009) and Bernalte et al. (2020) waters through surveillance monitoring programmes, employed a similar heavy metal detection device as a as foreseen in the WFD. On the other hand, the values monitoring tool for surface waters. The main differences obtained for the lowest detectable change (4–5 and were that the working electrode used in the study from 6–7 µg/L for Pb and Cu, respectively), indicate that Berho et al. (2009) was modified with the deposition of the developed integrated system is a promising tool a mercury salt, and the electrodes used in the Bernalte for investigative monitoring according to the WFD et al. (2020) study were composed of gold (working i.e. to ascertain the impact of urban and/or accidental and counter electrodes) and of silver (pseudo-reference pollution sources in inland surface waters. To confirm electrode). In both studies of Behro et al. (2009) and this hypothesis, different scenarios were calculated Bernalte et al. (2020), the instrument comprehended only by combining Pb and Cu emissions measured at cer- the SPE sensor coupled with a SWASV using a port- tain points of the urban water cycle, such as efflu- able device; however, they were not implemented on ent from conventional wastewater treatment plants an autonomous surface vehicle as in the present study. (WWTPs), from combined sewer overflow (CSO) The overall precision of the integrated system showed and stormwater runoff (Amann et al., 2019; Braun value ranges similar to the reproducibility calculated in et al. (2020); Clara et al., 2009; Clara et al., 2020; the study of Berho et al. (2009) (14–17% for Pb, 5–16% Clara et al., 2014; Clara et al., 2017; Hohenblum for Cu), but the limits of quantification obtained herein et al., 2000; Lambert et al., 2014; Revitt et al., 2019; were 5–7 times higher than the values reported in the Slobodnik et al., 2018; Toshovski et al., 2020; Zoboli earlier study (2 µg/L and 5 µg/L for Pb and Cu, respec- et al., 2019), and their possible dilutions into surface tively). Similarly, a high difference between the limit of water receivers of different sizes (BMLRT, 2002; quantification of the present study and of the study of Department of Environment Food and Rural Affairs Bernalte et al. (2020) can be observed (7.3 and 5.1 µg/L 2021; Wicke et al., 2015). For each studied heavy for Pb and Cu, respectively). This discrepancy can be metal (Cu and Pb), 15 scenarios were considered: explained by differences in the working electrodes, by three emission pathways (CSO, stormwater runoff, some differences among the SWASV set-ups and by WWTP) and five different urban situations for each the fact that some compromises had to be made during of them, such as large urban area and large receiver the integration of SWASV into the integrated system, (LL), large urban area and small receiver (LS), as discussed in “Validation through laboratory experi- medium urban area and medium receiver (MM), ments”. On the contrary, a comparison with the study medium urban area and small receiver (MS), small from Tasić et al. (2020) shows a similar or slightly bet- urban area and small receiver (SS). ter detection performance of Pb with the integrated sys- As seen in Fig. 9, the lowest detectable change tem (LOD of 6 µg/L and LOQ of 20 µg/L), and a less of the integrated system (dashed line) is suitable Vol:. (1234567890) 13 Environ Monit Assess (2022) 194:122 Page 15 of 19 122 high pollution emissions. Furthermore, it can be assumed that the validated system could be appro- priate to detect misconnections and cross connec- tions in urban areas through on-site and real-time monitoring of the receiving waters. These pollution sources are known to induce high immediate emis- sions into receiving water, but no calculation was possible for this case because of the lack of data in the literature regarding Pb and Cu concentra- tions emitted. From the different combinations of concentrations, surface waters and pathways, it was possible to observe that the integrated system is fit- for-purpose in detecting pollution plumes in surface water bodies when a large or medium urban area discharges into a small receiver. Conclusions An autonomous robotic system composed of a com- mercial autonomous surface vehicle equipped with additional features was developed to perform pre- programmed navigation in surface water bodies and simultaneous automatic sampling and chemical analysis. The architecture of the electronic system allows remote control of navigation, initialization of on-site measurements and collection of additional samples. The system is equipped with a microflu- idic device, which allows in-line sampling, filtering, Fig. 9  Potential concentrations of Cu and Pb in receiver water sample pretreatment and detection of heavy metals bodies for fifteen different scenarios: five urban area types (LL, using square wave anodic stripping voltammetry and LS, MM, MS, SS) for three output pathways (CSO, storm, WWTP). The dashed lines indicate the LDC values of the inte- screen-printed electrodes. The miniaturized dimen- grated system (6 µg/L and 4 µg/L for Cu and Pb, respectively) sions of components allow the implementation of an automated heavy metal detection device on a portable robotic system, enabling on-site investigative moni- for detection of Cu and Pb pollution coming from toring of source-point pollution in surface water. CSOs in large and medium urban areas combined As the calibration function strongly influences the with a small receiver, which can cause concentra- final measurement accuracy, its role was first evalu- tions up to 48 µg/L and 68 µg/L for Pb and Cu, ated. In the case of using SPE as a sensing element, respectively. The same conclusion can be reached this influence is of crucial importance; the calibra- for the detection of Cu pollution from CSOs and tion function is related to the specific electrode in use, WWTPs because Cu comes from large and medium which needs to be changed for each measuring cam- urban areas and is discharged in a small receiving paign. Three regression line trends were considered water body, and the majority of the calculated con- for the calibration function, and the forced-through- centration values are significantly higher than the zero linear calibration function was selected for fur- obtained LDC (concentrations up to 16 µg/L and ther calculations. 173 µg/L for CSOs and WWTPs, respectively). The lifetime of the sensor (maximal possible num- Detection of Cu pollution from stormwater runoff ber of measurements) was evaluated, as it was consid- can be considered possible but limited to extremely ered a limiting factor for field monitoring campaigns. Vol.: (0123456789) 13 122 Page 16 of 19 Environ Monit Assess (2022) 194:122 The average value of 39 measurements per sensor, this hypothesis, different scenarios for Pb and Cu which include a quality assurance test, the generation emissions in urban areas were calculated for water of a calibration function and effective on-site moni- body receivers of different sizes. The considered toring of the water body, was considered sufficient for scenarios included three emission pathways (CSO, the intended monitoring goals. stormwater runoff, WWTP) and five urban area An extensive validation of the integrated system, characteristics for each of them. It can be concluded partly based on ISO 15839 (International Standard, that the developed and validated autonomous robotic 2003, 2019), was performed to test its ability for system represents an innovative tool for investigative on-site detection of heavy metal pollution plumes monitoring of aquatic environment and is considered in river catchments. The main part of this valida- fit-for-purpose for the detection of Pb and Cu emis- tion process consisted of the simulation of pollution sions from large and medium urban areas that dis- plumes under laboratory conditions. Field experi- charge into a small water body receiver. ments were also performed at the Danube River to Further studies are ongoing to improve the configu- observe the behaviour and the performance of the ration and the performance characteristics of the devel- integrated system under field conditions. The fea- oped integrated system, aiming to decrease detection tures, autonomous navigation, continuous measure- limits, to improve the accuracy of multiple heavy metal ment of the water matrix with the HM microfluidic detection and to widen the range of metals detected. detection device and autonomous sampling, worked The development of such a portable autonomous moni- properly in concert, and all responded to the com- toring equipment is rather challenging, but a promising mands given through the GUI. The measurement alternative to standard laboratory methods. bias values, evaluated through recovery, were 75% and 65% for lead and copper, respectively (mean Acknowledgements The authors are grateful for the support and thankful for the opportunity to perform this research. values from 280 measurements). Although these values are not very satisfactory, especially com- Author contribution ED performed the laboratory and on- pared with those from reference laboratory analy- site experiments, contributed to their interpretation and the ses, they are considered acceptable for the purpose preparation of the manuscript. RA — as project leader for of this study, which is the detection of point pollu- BOKU University — managed the performance of the experi- ments, contributed to their interpretation, and contributed to tion (i.e. severe concentration change occurring in the manuscript preparation. WS and FZ assisted and supervised a short time) rather than accurate measurement of the performing of the laboratory and on-site experiments. They HM concentrations in surface water. The measure- performed the reference analysis and contributed to the data ment precision levels, evaluated as reproducibility interpretation. AF created the set-up of the autonomous surface vehicle, used for the validation experiments, and guaranteed and given as relative standard deviation (RSD in the needed technical support. QY, BN, RA and AM developed %), were 11–18% and 6–10% for Pb and Cu, respec- the heavy metal microfluidic detection device, which was used tively, which is considered satisfactory for on-site for the validation experiments. They guaranteed their technical measuring systems. The estimated limit of detection support during the experiments and contributed to the interpre- tation of the results. TK and AH contributed to the develop- (LOD) was rather poor (4 µg/L and 7 µg/L for Pb ment of the autonomous surface vehicle, by creating the online and Cu, respectively) for monitoring and assessing cloud storage of data and the specific software needed for this the chemical status of surface waters through sur- application. All the authors reviewed the manuscript. veillance monitoring programmes, according to the European Water Framework Directive. As the low- Funding Open access funding provided by University of Natural Resources and Life Sciences Vienna (BOKU). The est detectable change (LDC) is considered the most research leading to the presented results received funding from important performance characteristic for the detec- the research project INTCATCH 2020, “Development and tion of pollution plumes in a water body, intensive application of Novel, Integrated Tools for monitoring and man- experiments were carried out to estimate this per- aging Catchments” supported by the European Union’s Hori- zon 2020 research and innovation programme, under the grant formance characteristic. The estimated values for agreement No. 689341. The ICN2 is funded by the CERCA the lowest detectable change were 4–5 µg/L and Programme/Generalitat de Catalunya. The ICN2 is supported 6–7 µg/L for Pb and Cu, respectively, which are con- by the Severo Ochoa program of the Spanish Ministry of sidered satisfactory for detecting severe concentra- Economy, Industry, and Competitiveness (MINECO, Grant No. SEV-2017–0706). tion gradients in a specific water body. To support Vol:. (1234567890) 13 Environ Monit Assess (2022) 194:122 Page 17 of 19 122 Availability of data and materials The datasets used and/or Chemical, 307, 127620. https://​doi.​org/​10.​1016/j.​snb.​ analysed during the current study are available from the cor- 2019.​127620 responding author on reasonable request. BMLRT. (2002). eHYD Hydrographic yearbook interactive map. https://​ehyd.​gv.​at/#. Accessed 24 March 2021. Code availability Not applicable. Boggs Jr, S., Livermore, D., & Seitz, M. G. (1985). Humic substances in natural waters and their complexation with trace metals and radionuclides: a review. [129 references]. Declarations (Report ANL-84-78, ON: DE85015539). United States. https://​doi.​org/​10.​2172/​55699​09 Borrill, A. J., Reily, N. E., & Macpherson, J. V. (2019). Address- Ethics approval Not applicable. ing the practicalities of anodic stripping voltammetry for heavy metal detection: A tutorial review. The Analyst, 144, Consent to participate Not applicable. 6834–6849. https://​doi.​org/​10.​1039/​C9AN0​1437C Braun, R., Hartmann, C., Kreuzinger, N., Lenz, K., Schaar, H., & Scheffknecht, C. (2020). Untersuchung von Abwässern Consent for publication Not applicable. und Gewässern auf unterschiedliche toxikologische End- punkte: Biologische Wirktests mittels in-vitro-Verfahren. Competing interests The authors declare no competing inter- Butterfield, M. (2009). Monitoring for heavy metals. Air, ests. Water, Environment International (57). Clara, M., Denner, M., Gans, O., Scharf, S., Windhofer, G., & Zessner, M. (2009). Emissionen organischer und anorga- Open Access This article is licensed under a Creative Com- nischer Stoffe aus kommunalen Kläranlagen. (Report REP- mons Attribution 4.0 International License, which permits 0247,). Wien. http://​www.​umwel​tbund​esamt.​at/​filea​dmin/​ use, sharing, adaptation, distribution and reproduction in any site/​publi​katio​nen/​REP02​47.​pdf medium or format, as long as you give appropriate credit to the Clara, M., Gruber, G., Hohenblum, P., Hofer, T., Kittlaus, S., original author(s) and the source, provide a link to the Crea- Lenz, K., et al. (2020). TEMPEST: Erfassung von Emis- tive Commons licence, and indicate if changes were made. The sionen ausgewählter Spurenstoffe aus Kanalsystemen, images or other third party material in this article are included Handlungsoptionen zu deren Minderung und Opimierung in the article’s Creative Commons licence, unless indicated oth- einer alternativen Nachweismethode für Kunststoffpar- erwise in a credit line to the material. If material is not included tikel in Wasserproben. Wien. in the article’s Creative Commons licence and your intended Clara, M., Gruber, G., Humer, F., Hofer, T., Kretschmer, F., Ertl, T., use is not permitted by statutory regulation or exceeds the per- et al. (2014). Spurenstoffemissionen aus Siedlungsgebieten und mitted use, you will need to obtain permission directly from the von Verkehrsflächen. SCHTURM. Wien. https://​www.​bmnt.​ copyright holder. To view a copy of this licence, visit http://​ gv.​at/​servi​ce/​publi​katio​nen/​wasser/​Spure​nstof​f emis​sionen-​aus-​ creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. Siedl​ungsg​ebiet​en-​und-​von-​Verke​hrsfl​aechen.​html Clara, M., Hanefeld, W., & Scheffknecht, C. (2017). Unter- suchung ausgewählter prioritärer und sonstiger Stoffe in kommunalen Kläranlagen und Fließge-wässern in Vorar- lberg Fließgewässern in Vorarlberg. https://​vorar​lberg.​at/​ References web/​land-​vorar​lberg/​conte​ntdet​ailse​ite/-/​asset_​publi​sher/​ qA6AJ​38txu​0k/​conte​nt/​prior​itaere-​stoffe-​in-​kommu​nalen-​ Amann, A., Clara, M., Gabriel, O., Hochedlinger, G., Humer, klaer​anlag​en-​und-​flies​sgewa​essern?​artic​le_​id=​117972 M., Humer, F., Kittlaus, S., Kulcsar, S., Scheffknecht, de la Escosura-Muñiz, A., Ambrosi, A., & Merkoçi, A. (2008). C., Trautvetter, H., Zessner, M., & Zoboli, O. (2019). Electrochemical analysis with nanoparticle-based biosys- STOBIMO Spurenstoffe, Stoffbilanzmodellierung für tems. TrAC Trends in Analytical Chemistry, 27, 568–584. Spurenstoffe auf Einzugsgebietsebene. Wien. https://​doi.​org/​10.​1016/j.​trac.​2008.​05.​008 Barón-Jaimez, J., Joya, M. R., & Barba-Ortega, J. (2013). Department of Environment Food and Rural Affairs. (2021). Anodic stripping voltammetry ASV for determination of Hydrology data explorer. Department of Environment Food heavy metals. Journal of Physics: Conference Series, 466, and Rural Affairs. http://​envir​onment.​data.​gov.​uk/​hydro​logy/​ 12023. stati​on/​8496c​e69-​482c-​406a-​a2f0-​ac418​ef8f0​99 Berho, C., Guigues, N., Ghestem, J. P., Crouzet, C., Strugeon, EEA. (1996). European freshwater monitoring network design A., Roy, S., et al. (2009). On-site heavy metal monitor- (10/1996). ing using a portable screen-printed electrode sensor. In C. EEA. (2018). Environmental pressure of heavy metal releases from Gonzalez, R. Greenwood, & P. Quevauviller (Eds.), Rapid Europe’s industry (Industrial pollution in Europe No 3/2018). chemical and biological techniques for water monitoring https://​www.​eea.​europa.​eu/​publi​catio​ns/​envir​onmen​tal-​ (pp. 263–273, Water quality measurements series). Chich- press​ures-​of-​heavy-​metal. Accessed 17 July 2020. ester, U.K.: Wiley. EEA, Whalley, C., Mohaupt, V., Busch, W., van den Roovart, Bernalte, E., Arévalo, S., Pérez-Taborda, J., Wenk, J., Estrela, J., van Dujnhoven, N., et al. (2018). Chemicals in Euro- P., Avila, A., et al. (2020). Rapid and on-site simultane- pean waters: Knowledge developments (18/2018). ous electrochemical detection of copper, lead and mer- EU. (2000). Directive 2000/60/EC of the European Parlia- cury in the Amazon river. Sensors and Actuators B: ment and of the Council of 23 October 2000 establishing Vol.: (0123456789) 13 122 Page 18 of 19 Environ Monit Assess (2022) 194:122 a framework for community action in the field of water für den Eintrag von Schadstoffen aus kommunalen policy: WFD. Kläranlagen in Gewässer. Abschlussbericht gefördert EU. (2013). Directive 2013/39/EU of the European Parliament unter dem Az: 29630 von der Deutschen. https://​www.​ and of the Council of 12 August 2013 amending Direc- dbu.​de/​OPAC/​ab/​DBU-​Absch​lussb​ericht-​AZ-​29630.​pdf. tives 2000/60/EC and 2008/105/EC as regards priority Accessed 6 May 2020. substances in the field of water policy Text with EEA rel- Law, C. S., Lim, S. Y., Abell, A. D., & Santos, A. (2018). evance: WFD. Real-time binding monitoring between human blood Gautam, R., Sharma, S., Mahiya, S., & Chattopadhyaya, M. proteins and heavy metal ions in nanoporous anodic alu- (2014). Heavy metals in water: Presence, removal and safety. mina photonic crystals. Analytical Chemistry, 90, 10039– Gromaire, M. C., GARNAUD, S., Saad, M., & Chebbo, G. 10048. https://​doi.​org/​10.​1021/​acs.​analc​hem.​8b027​32 (2001). Contribution of different sources to the pollu- Li, M., Cao, R., Nilghaz, A., Guan, L., Zhang, X., & Shen, W. tion of wet weather flows in combined sewers. Water (2015). “Periodic-table-style” paper device for monitoring Research, 35(2), 521–533. heavy metals in water. Analytical Chemistry, 87, 2555– Güell, R., Aragay, G., Fontàs, C., Anticó, E., & Merkoçi, A. 2559. https://​doi.​org/​10.​1021/​acs.​analc​hem.​5b000​40 (2008). Sensitive and stable monitoring of lead and cad- Lu, Y., Liang, X., Niyungeko, C., Zhou, J., Xu, J., & Tian, G. mium in seawater using screen-printed electrode and elec- (2018). A review of the identification and detection of heavy trochemical stripping analysis. Analytica Chimica Acta, metal ions in the environment by voltammetry. Talanta, 178, 627, 219–224. https://​doi.​org/​10.​1016/j.​aca.​2008.​08.​017 324–338. https://​doi.​org/​10.​1016/j.​talan​ta.​2017.​08.​033 Gumpu, M. B., Sethuraman, S., Krishnan, U. M., & Rayappan, J. Masindi, V., & Muedi, K. L. (2018). Environmental contamina- B. B. (2015). A review on detection of heavy metal ions in tion by heavy metals: 7. In H. E.-D. M. Saleh & R. F. Aglan water – An electrochemical approach. Sensors and Actua- (Eds.), Heavy metals (pp. 115–133). Rijeka: IntechOpen. tors B: Chemical, 213, 515–533. https://​doi.​org/​10.​1016/j.​ Melo, M., Mota, F., Albuquerque, V., & Alexandria, A. (2019). snb.​2015.​02.​122 Development of a Robotic Airboat for Online Water Qual- HACH. (2019). Copper trace cuvette test. https://​uk.​hach.​com/​ ity Monitoring in Lakes (Robotics, Vol. 8). copper-​trace-​cuvet​te-​test-0-​01-1-​0-​mg-l-​cu-​20-​tests/​produ​ct-​ Merkoçi, A., Vasjari, M., Fàbregas, E., & Alegret, S. (2000). detai​ls?​id=​26370​291498. Accessed 9 January 2020. Determination of Pb and Cu by anodic stripping voltam- HACH. (2020). Lead cuvette test. https://​uk.​hach.​com/​lead-​ metry using glassy carbon electrodes modified with mer- cuvet​te-​test-0-​1-2-​0-​mg-l-​pb-​25-​tests/​produ​ct-​downl​oads?​ cury or mercury-nafion films. Microchimica Acta, 135, id=​26370​29140​2&​callb​ack=​qs. Accessed 9 January 2020. 29–33. https://​doi.​org/​10.​1007/​s0060​40070​015 Hohenblum, P., Sattelberger, R., & Scharf, S. (2000). ABWAS- Mohammed, N., Baidya, A., Murugesan, V., Kumar, A. A., SER- UND KLÄRSCHLAMMUNTERSUCHUNGEN IN Ganayee, M. A., Mohanty, J. S., et al. (2016). Diffusion- DER PILOTKLÄRANLAGEENTSORGUNGSBETRIEBE controlled simultaneous sensing and scavenging of heavy SIMMERING (EbS) (M-121). Umweltbundesamt GmbH. metal ions in water using atomically precise cluster– International Standard. (2003, 2019). Water quality - On-line sen- cellulose nanocrystal composites. ACS Sustainable sors/analysing equipment for water: Specifications and perfor- Chemistry & Engineering, 4, 6167–6176. https://​doi.​org/​ mance tests (13.060.01, 13.060.45). Switzerland, 13.060.01, 10.​1021/​acssu​schem​eng.​6b016​74 13.060.45(15839). https://​www.​iso.​org/​stand​ard/​28740.​html Mostofa, K., Liu, C. Q., Mottaleb, M. A., Wan, G., Ogawa, H., International Standard. (2012). Water quality - estimation of Vione, D., et al. (2013a). Dissolved organic matter in nat- measurement uncertainty based on validation and qual- ural waters. In (pp. 1–137). ity control data (Vol. 13.060.45), 13.060.45(11352:2012). Mostofa, K. M. G., Liu, C. Q., Feng, X., Yoshioka, T., Vione, D., https://​www.​iso.​org/​stand​ard/​50399.​html Pan, X., et al. (2013b). Complexation of dissolved organic Jang, A., Zou, Z., Lee, K. K., Chong, A. H., & Bishop, P. L. matter with trace metal ions in natural waters. In K. M. G. (2011). State-of-the-art lab chip sensors for environmen- Mostofa (Ed.), Photobiogeochemistry of organic matter: tal water monitoring. Measurement Science and Tech- Principles and practices in water environments (Vol. 55, pp. nology, 22, 32001. https://​doi.​org/​10.​1088/​0957-​0233/​ 769–849, Environmental science and engineering. Environ- 22/3/​032001 mental science). Heidelberg, New York: Springer. Khan, Y., Thielens, A., Muin, S., Ting, J., Baumbauer, C., & Mostofa, K. M. G. (Ed.). (2013). Photobiogeochemistry of Arias, A. C. (2020). A new frontier of printed electron- organic matter: Principles and practices in water environ- ics: Flexible hybrid electronics. Advanced Materials, ments (Environmental science and engineering. Environ- 32, 1905279. https://​doi.​org/​10.​1002/​adma.​20190​5279 mental science). Springer. Koller, M., & Saleh, H. M. (Eds.). (2018). Introductory chap- Müller, A., Österlund, H., Nordqvist, K., Marsalek, J., & ter: Introducing heavy metals. United Kingdom, Lon- Viklander, M. (2019). Building surface materials as don: IntechOpen. sources of micropollutants in building runoff: A pilot Lam, M. T., Chakrabarti, C. L., Cheng, J., & Pavski, V. study. The Science of the Total Environment, 680, 190– (1997). Rotating disk electrode voltammetry/anodic 197. https://​doi.​org/​10.​1016/j.​scito​tenv.​2019.​05.​088 stripping voltammetry for chemical speciation of lead Odobašić, A., Šestan, I., & Begić, S. (Eds.). (2019). Biosensors and cadmium in freshwaters containing dissolved for determination of heavy metals in waters. organic matter. Electroanalysis, 9, 1018–1029. https://​ Palchetti, I., Laschi, S., & Mascini, M. (2005). Miniaturised doi.​org/​10.​1002/​elan.​11400​91310 stripping-based carbon modified sensor for in field analy- Lambert, B., Fuchs, S., Toshovski, S., Sacher, F., & Thoma, sis of heavy metals. Analytica Chimica Acta, 530, 61–67. A. (2014). Entwicklung eines Bilanzierungsinstruments https://​doi.​org/​10.​1016/j.​aca.​2004.​08.​065 Vol:. (1234567890) 13 Environ Monit Assess (2022) 194:122 Page 19 of 19 122 Pandey, A. K., Dhar Pandey, S., & Misra, V. (2000). Stability con- ions in biology, biochemistry, and medicine, 18, 121–129. stants of metal–humic acid complexes and its role in envi- https://​doi.​org/​10.​1007/​s10534-​004-​5787-3. ronmental detoxification. Ecotoxicology and Environmental Waheed, A., Mansha, M., & Ullah, N. (2018). Nanomaterials- Safety, 47, 195–200. https://​doi.​org/​10.​1006/​eesa.​2000.​1947 based electrochemical detection of heavy metals in water: Pisarek, I., & Głowacki, M. (2015). Quality of groundwater Current status, challenges and future direction. TrAC and aquatic humic substances from main reservoire of Trends in Analytical Chemistry, 105, 37–51. https://​doi.​ Ground Water No. 333. Journal of Ecological Engineer- org/​10.​1016/j.​trac.​2018.​04.​012 ing, 16, 46–53. https://​doi.​org/​10.​12911/​22998​993/​60453 Wang, N., Kanhere, E., Kottapalli, A. G. P., Miao, J., & Triantafyllou, Rensing, C., & Maier, R. M. (2003). Issues underlying use of M. S. (2017). Flexible liquid crystal polymer-based electro- biosensors to measure metal bioavailability. Ecotoxicology chemical sensor for in-situ detection of zinc(II) in seawater. and Environmental Safety, 56, 140–147. https://​doi.​org/​ Microchimica Acta, 184, 3007–3015. https://​doi.​org/​10.​1007/​ 10.​1016/​S0147-​6513(03)​00057-5 s00604-​017-​2280-6 Revitt, D. M., Ellis, J. B., & Lundy, L. (2019). Justification of Wicke, D., Matzinger, A., & Rouault P. (2015). Relevanz orga- an excel spreadsheet approach for predicting road runoff nischer Spurenstoffe im Regenwasserabfluss Berlins, pollutant concentrations. Accessed 29 March 2021. Wiklund, J., Karakoç, A., Palko, T., Yiğitler, H., Ruttik, K., Jäntti, Silva Junior, G. A., Lima Sa, T. S., Santos, H. D., Negreiros, P. R., et al. (2021). A review on printed electronics: Fabrica- Á., Souza Silva, M. J., Álvarez Jácobo, E. J., et al. (2016). tion methods, inks, substrates, applications and environmen- Towards a real-time embedded system for water monitor- tal impacts. Journal of Manufacturing and Materials Pro- ing installed in a robotic sailboat (Sensors, Vol. 16). cessing. https://​doi.​org/​10.​3390/​jmmp5​030089 Slobodnik, J., Brack, W., van Gils, J., Dulio, V., & Liska, I. Winters, N., Granuke, K., & McCall, M. (2015). Roofing materi- (2018). Guidance for identification of RBSPs and list of als assessment: Investigation of five metals in runoff from Danube RBSPs including quantification of their ecologi- roofing materials. Water Environment Research: A Research cal impact and modelling-based exposure and risk pre- Publication of the Water Environment Federation, 87, 835– dictions validated with case-study data. EU SOLUTIONS 848. https://​doi.​org/​10.​2175/​10614​3015X​14362​86522​6437 Project. External Deliverable, 19, 4. Winters, N. L., & Graunke, K. (2014). Roofing materials Steccanella, L., Bloisi, D., Blum, J., & Farinelli, A. (2019). assessment: Investigation of toxic chemicals in roof run- Deep learning waterline detection for low-cost autono- off. Washington State Department of Ecology Olympia, mous boats. In M. Strand, R. Dillmann, E. Menegatti, & Washington, 98504–7710. S. Ghidoni (Eds.) (pp. 613–625). Cham: Springer Interna- Yang, Q., Nagar, B., Alvarez-Diduk, R., Balsells, M., Farinelli, tional Publishing. A., Bloisi, D., et al. (2021). Development of a heavy metal Steinberg, C., Paul, A., Pflugmacher, S., Meinelt, T., Klöcking, sensing boat for automatic analysis in natural waters uti- R., & Wiegand, C. (2003). Pure humic substances have lizing anodic stripping voltammetry. ACS ES&T Water. the potential to act as xenobiotic chemicals - A review. https://​doi.​org/​10.​1021/​acses​twater.​1c001​92 Fresenius Environmental Bulletin, 12, 391–401. Zinoubi, K., Majdoub, H., Barhoumi, H., Boufi, S., Jaffrezic- Tasić, N., de Sousa Oliveira, L., Paixão, T. R. L. C., & Moreira Renault, N. (None). (2017). Determination of trace heavy Gonçalves, L. (2020). Laser-pyrolysed paper electrodes metal ions by anodic stripping voltammetry using nanofi- for the square-wave anodic stripping voltammetric detec- brillated cellulose modified electrode. Journal of Elec- tion of lead. MEDICAL DEVICES & SENSORS, 3, troanalytical Chemistry, 799, 70-77. https://​doi.​org/​10.​ e10115. https://​doi.​org/​10.​1002/​mds3.​10115 1016/j.​jelec​hem.​2017.​05.​039 Toshovski, S., Kaiser, M., Fuchs, S., Sacher, F., Thoma, Zoboli, O., Clara, M., Gabriel, O., Scheffknecht, C., Humer, A., Kümmel, V., et al. (2020). Prioritäre Stoffe in M., Brielmann, H., et al. (2019). Occurrence and levels kommunalen Kläranlagen: Ein deutschlandweit har- of micropollutants across environmental and engineered monisiertes Vorgehen (173). Dessau-Roßlau. https://​ compartments in Austria. Journal of Environmental www.​u mwel​t bund​e samt.​d e/​p ubli​k atio​n en/​p rior ​i taere-​ Management, 232, 636–653. https://​doi.​org/​10.​1016/j.​ stoffe-​in-​kommu​nalen-​klaer​anlag​en jenvm​an.​2018.​10.​074 Tuna, G., Arkoc, O., & Gulez, K. (2013). Continuous moni- toring of water quality using portable and low-cost Publisher’s Note Springer Nature remains neutral with regard approaches. International Journal of Distributed Sensor to jurisdictional claims in published maps and institutional Networks, 9, 249598. https://​doi.​org/​10.​1155/​2013/​249598 affiliations. Verma, N., & Singh, M. (2005). Biosensors for heavy metals. Biometals: An international journal on the role of metal Vol.: (0123456789) 13