Search

Traditional laboratory-based water quality monitoring and testing approaches are soon to be outdated, mainly because of the need for real-time feedback and immediate responses to emergencies. The more recent wireless sensor network (WSN)-based techniques are evolving to alleviate the problems of monitoring, coverage, and energy management, among others. The inclusion of the Internet of Things (IoT) in WSN techniques can further lead to their improvement in delivering, in real time, effective and efficient water-monitoring systems, reaping from the benefits of IoT wireless systems. However, they still suffer from the inability to deliver accurate real-time data, a lack of reconfigurability, the need to be deployed in ad hoc harsh environments, and their limited acceptability within industry. Electronic sensors are required for them to be effectively incorporated into the IoT WSN water-quality-monitoring system. Very few electronic sensors exist for parameter measurement. This necessitates the incorporation of artificial intelligence (AI) sensory techniques for smart water-quality-monitoring systems for indicators without actual electronic sensors by relating with available sensor data. This approach is in its infancy and is still not yet accepted nor standardized by the industry. This work presents a smart water-quality-monitoring framework featuring an intelligent IoT WSN monitoring system. The system uses AI sensors for indicators without electronic sensors, as the design of electronic sensors is lagging behind monitoring systems. In particular, machine learning algorithms are used to predict E. coli concentrations in water. Six different machine learning models (ridge regression, random forest regressor, stochastic gradient boosting, support vector machine, k-nearest neighbors, and AdaBoost regressor) are used on a sourced dataset. From the results, the best-performing model on average during testing was the AdaBoost regressor (a M A E ¯ of 14.37 counts/100 mL), and the worst-performing model was stochastic gradient boosting (a M A E ¯ of 42.27 counts/100 mL). The development and application of such a system is not trivial. The best-performing water parameter set (Set A) contained pH, conductivity, chloride, turbidity, nitrates, and chlorophyll. [ABSTRACT FROM AUTHOR]

Sodium hypochlorite is a widely used additive in water used to disinfect and remove any disease‐causing bacteria that can be found in sources of water, and is used to wash contaminants from meats, fruits, and vegetables. Many free chlorine sensors exist which monitor free chlorine levels such as the use of colorimetric or electrochemical methods, ensuring that the free chlorine is within safe and regulated levels. However colorimetric sensing methods are irreversible and often use toxic compounds, and electrochemical sensors, although reversible, are made with materials which are not suitable to be used near drinking water or food products. By developing sensors which are made using alternative materials and methods, the sensors can be used in and around drinking water and food products. This review article discusses various electrochemical‐free chlorine sensors made with various carbon‐based materials and methods resulting in sensors that are biodegradable, relatively inert, and resilient in the presence of harsh chemicals, making them safe to use near and around food and still maintain competitive performance parameters. This review article showcases some of the recent progress, the importance, preconditions, and the various future needs and potentials of carbon‐based electrochemical‐free chlorine sensors. [ABSTRACT FROM AUTHOR]

Traditional laboratory-based water quality monitoring and testing approaches are soon to be outdated, mainly because of the need for real-time feedback and immediate responses to emergencies. The more recent wireless sensor network (WSN)-based techniques are evolving to alleviate the problems of monitoring, coverage, and energy management, among others. The inclusion of the Internet of Things (IoT) in WSN techniques can further lead to their improvement in delivering, in real time, effective and efficient water-monitoring systems, reaping from the benefits of IoT wireless systems. However, they still suffer from the inability to deliver accurate real-time data, a lack of reconfigurability, the need to be deployed in ad hoc harsh environments, and their limited acceptability within industry. Electronic sensors are required for them to be effectively incorporated into the IoT WSN water-quality-monitoring system. Very few electronic sensors exist for parameter measurement. This necessitates the incorporation of artificial intelligence (AI) sensory techniques for smart water-quality-monitoring systems for indicators without actual electronic sensors by relating with available sensor data. This approach is in its infancy and is still not yet accepted nor standardized by the industry. This work presents a smart water-quality-monitoring framework featuring an intelligent IoT WSN monitoring system. The system uses AI sensors for indicators without electronic sensors, as the design of electronic sensors is lagging behind monitoring systems. In particular, machine learning algorithms are used to predict E. coli concentrations in water. Six different machine learning models (ridge regression, random forest regressor, stochastic gradient boosting, support vector machine, k-nearest neighbors, and AdaBoost regressor) are used on a sourced dataset. From the results, the best-performing model on average during testing was the AdaBoost regressor (a MAE¯ of 14.37 counts/100 mL), and the worst-performing model was stochastic gradient boosting (a MAE¯ of 42.27 counts/100 mL). The development and application of such a system is not trivial. The best-performing water parameter set (Set A) contained pH, conductivity, chloride, turbidity, nitrates, and chlorophyll.

Water systems are rapidly transforming into cyber-physical systems. Despite the benefits of remote control and monitoring, autonomous operation and connectivity, there is an expanded threat surface, which includes cyber-physical attacks. This study demonstrates a stress-testing methodology that focuses on assessing the performance of a contamination warning system, designed with alternative water quality (WQ) sensor placement strategies against cyber-physical attacks. The physical part of the attacks consists of backflow injection attacks with a contaminant, while the cyber part comprises cyber-attacks to the contamination warning system. The WQ sensor designs are generated with the Threat Ensemble Vulnerability Assessment and Sensor Placement Optimization Tool (TEVA-SPOT), based on optimizing various metrics. The coupled WDN and CPS operation, the deliberate contamination events, and the cyber-physical attacks, are simulated with the water system cyber-physical stress-testing platform RISKNOUGHT. Multidimensional resilience profile graphs are utilized to analyze performance, demonstrated in a benchmark case study. This type of assessment can be useful in risk assessment studies for water utilities as well as in WQ sensor placement optimization.

Water distribution network (WDN) is one of the most essential infrastructures all over the world and ensuring water quality is always a top priority. To this end, water quality sensors are often deployed at multiple points of WDNs for real-time contamination detection and fast contamination source identification (CSI). Specifically, CSI aims to identify the location of the contamination source, together with some other variables such as the starting time and the duration. Such information is important in making an efficient plan to mitigate the contamination event. In the literature, simulation-optimisation methods, which combine simulation tools with evolutionary algorithms (EAs), show great potential in solving CSI problems. However, the application of EAs for CSI is still facing big challenges due to their high computational cost. In this paper, we propose DLGEA, a deep learning guided evolutionary algorithm to improve the efficiency by optimising the search space of EAs. Firstly, based on a large number of simulated contamination events, DLGEA trains a deep neural network (DNN) model to capture the relationship between the time series of sensor data and the contamination source nodes. Secondly, given a contamination event, DLGEA guides the initialisation and optimise the search space of EAs based on the top K contamination nodes predicated by the DNN model. Empirically, based on two benchmark WDNs, we show that DLGEA outperforms the CSI method purely based on EAs in terms of both the average topological distance and the accumulated errors between the predicted and the real contamination events. [ABSTRACT FROM AUTHOR]

This paper aims to assess and analyze the health impact of consuming contaminated drinking water in a water distributed system (WDS). The analysis was based on qualitative simulation performed in two different models named hydraulic and water quality in a WDS. The computation focuses on quantitative analysis for chemically contaminated water impacts by analyzing the dose level in various locations in the water network and the mass of the substance that entered the human body. Several numerical experiments have been applied to evaluate the impact of water pollution on human life. They analyzed the impact on human life according to various factors, including the location of the injected node (pollution occurrence) and the ingested dose level. The results show a significant impact of water contaminant on human life in multiple areas in the water network, and the level of this impact changed from one location to another in WDSs based on several factors such as the location of the pollution occurrence, the contaminant concentration, and the dose level. In order to reduce the impact of this contaminant, water quality sensors have been used and deployed on the water network to help detect this contaminant. The sensors were optimally deployed based on the time-detection of water contamination and the volume of polluted water consumed. Numerical experiments were carried out to compare water pollution's impact with and without using water quality sensors. The results show that the health impact was reduced by up to 98.37% by using water quality sensors. [ABSTRACT FROM AUTHOR]

Abstract Automation and reliability are the crucial elements of any advance reverse osmosis plant to meet the environmental and economic demands. Early fault indication, diagnosis and regular maintenance are the key challenges with most of the reverse osmosis plants in the Indian scenario. The present work introduces a modern reverse osmosis (RO) plant status monitoring unit to monitor different plant parameters in real time and early prediction for faults and maintenance. Developed RO plant status monitoring unit consists of a touch screen-based embedded monitoring unit, water quality sensors (pH, TDS), sampling chamber for controlled water flow, flow sensors, pressure and level sensors. The present system has been developed in a modular fashion so that it could be integrated with any capacity of RO plant units. Developed embedded system monitors various parameters of the plant such as input power, efficiency of the plant, level of input and output water tank and also guides operator with instructions for plant operation. Other than this, a dedicated smartphone app interface has been developed for the operator to acquire data from status monitoring unit, storage on smartphone, and transfer it to the cloud. The developed smartphone-based app also provides facility to integrate plant data with Google map with location information for easy understanding and quick action. The system has also a backup facility to transfer data to the server using 2G GSM module during the unavailability of the operator. A dedicated centralized Web server has been developed for real-time visualization of all installed RO plant status monitoring units. Different machine learning techniques have been implemented on acquired sensors data to predict early warnings related to power failure, membrane fouling and scaling, input water shortage, pipe, tank leakage, water quality sensors damage, non-operation or wrong operation of the plant along with different maintenance actions such as membrane water and chemical wash. Developed RO status monitoring unit has been tested with various RO plants having capacity from 500 LPH to 2000 LPH and deployed at various nearby villages of Rajasthan.

Abstract Water quality in rural areas is difficult to monitor due to lack of connectivity from different water laboratories. In other areas, location-based real-time water quality data collection is a tedious job and highly dependent on human intervention. The presented paper introduces a low-cost battery operated smartphone-based embedded system design to measure different water quality parameters in various remote locations. Developed system measures pH, total dissolved salt (TDS) and temperature of the water samples using off the shelf available sensors. Measured pH and TDS dataset have been used to derive other water quality parameters using standard mathematical relationships such as salinity, oxygen reduction potential and conductivity. Front-end readout interface circuit has been designed and interfaced with 8-bit microcontroller along with classical Bluetooth module for measurement, data acquisition, and logging purpose. A dedicated smartphone-based application offers analysis and cloud data storage possibilities. It also provides facility to analyze water quality data with location information on Google map for quick judgment and easy understanding. The developed smartphone-based application provides the facility of auto-calibration feature for rapid and on-site usage. Developed smartphone-based application also opens up the possibility to share the data and warnings using different options such as SMS, WhatsApp and E-mail. Overall device has dimensions of 11.0 × 8.0 × 4.0 (in cm), weighs 350 g and runs with 9-V rechargeable battery. Obtained results have been validated with standard water quality measurement system from Eutech Instruments, and it has been observed that measured and calculated parameters are acceptable according to Indian water quality standards. Various statistical and artificial neural network-based modeling techniques have been used to convert measured water quality parameters to a single water quality index for easy and rapid judgment. The developed water quality measurement system has been used for multiple applications to explore the utility of the system such as instant water quality judgment and real-time water quality analysis of different water sources. One of the other explored applications is the real-time water quality monitoring of small ponds and lakes.

Different types of sensors are available in various ranges in size and design for use in the aquaculture industry. Recent research and development in designing of various types of water quality sensors and biosensors for monitoring and controlling important aquaculture parameters opens a new avenue to make the aquaculture system sustainable. The sensor to be used in aquaculture should fulfil certain specific designs to meet underwater operational requirements. The comprehensive investigation for performance assessment of existing sensors is necessary to achieve better outcomes. The comparative interpretation would provide the relevant information to the sensor designer and aqua farmers. Literature review on existing sensors reveals that the studies for performance evaluation on existing sensors is not abridged. Therefore, it becomes quite difficult for selection of a particular sensor in aquaculture. This review paper summarized different architypes developed for designing sensors for deployment in aquaculture purposes. The additional parameters have been given in the paper for improvement of designing of sensors. Based on this review paper, a procedure has been proposed for performance evaluation of sensors used in aquaculture. This would help in designing and selection of sensors in the aquaculture industry. [ABSTRACT FROM AUTHOR]

This paper aims to assess and analyze the health impact of consuming contaminated drinking water in a water distributed system (WDS). The analysis was based on qualitative simulation performed in two different models named hydraulic and water quality in a WDS. The computation focuses on quantitative analysis for chemically contaminated water impacts by analyzing the dose level in various locations in the water network and the mass of the substance that entered the human body. Several numerical experiments have been applied to evaluate the impact of water pollution on human life. They analyzed the impact on human life according to various factors, including the location of the injected node (pollution occurrence) and the ingested dose level. The results show a significant impact of water contaminant on human life in multiple areas in the water network, and the level of this impact changed from one location to another in WDSs based on several factors such as the location of the pollution occurrence, the contaminant concentration, and the dose level. In order to reduce the impact of this contaminant, water quality sensors have been used and deployed on the water network to help detect this contaminant. The sensors were optimally deployed based on the time-detection of water contamination and the volume of polluted water consumed. Numerical experiments were carried out to compare water pollution’s impact with and without using water quality sensors. The results show that the health impact was reduced by up to 98.37% by using water quality sensors.

Diatoms are among the dominant phytoplankters in marine and freshwater habitats, and important biomarkers of water quality, making their identification and classification one of the current challenges for environmental monitoring. To date, taxonomy of the species populating a water column is still conducted by marine biologists on the basis of their own experience. On the other hand, deep learning is recognized as the elective technique for solving image classification problems. However, a large amount of training data is usually needed, thus requiring the synthetic enlargement of the dataset through data augmentation. In the case of microalgae, the large variety of species that populate the marine environments makes it arduous to perform an exhaustive training that considers all the possible classes. However, commercial test slides containing one diatom element per class fixed in between two glasses are available on the market. These are usually prepared by expert diatomists for taxonomy purposes, thus constituting libraries of the populations that can be found in oceans. Here we show that such test slides are very useful for training accurate deep Convolutional Neural Networks (CNNs). We demonstrate the successful classification of diatoms based on a proper CNNs ensemble and a fully augmented dataset, i.e., creation starting from one single image per class available from a commercial glass slide containing 50 fixed species in a dry setting. This approach avoids the time-consuming steps of water sampling and labeling by skilled marine biologists. To accomplish this goal, we exploit the holographic imaging modality, which permits the accessing of a quantitative phase-contrast maps and a posteriori flexible refocusing due to its intrinsic 3D imaging capability. The network model is then validated by using holographic recordings of live diatoms imaged in water samples i.e., in their natural wet environmental condition.

A water quality model (WQM) is a tool used for the prediction of water quality in a water distribution network. Previous studies have proposed a variety of methods for computing WQMs. All of these are flow-based and thus require the use of a hydraulic model. The current paper proposes a new methodology for a WQM based on a statistical approach. The study suggests a methodology that predicts the change in both detention time and water quality between two locations in a branched network, without the use of a hydraulic model. The methodology was implemented at a large water utility and tested using real-time data. The model output was then examined and verified by conducting a statistical test to compare the detention time with that predicted using an existing hydraulic model. It was found that the proposed model can be effectively used as a WQM. [ABSTRACT FROM AUTHOR]

Abstract: Streams provide a physical linkage between land and downstream river networks, delivering solutes derived from multiple catchment sources. We analyzed high‐frequency time series of stream solutes to characterize the timing and magnitude of major ion, nutrient, and organic matter transport over event, seasonal, and annual timescales as well as to assess whether nitrate ( NO 3 −) and dissolved organic carbon (DOC) transport are coupled in catchments, which would be expected if they are subject to similar biogeochemical controls throughout the watershed. Our data set includes in situ observations of NO 3 −, fluorescent dissolved organic matter (DOC proxy), and specific conductance spanning 2–4 years in 10 streams and rivers across New Hampshire, including observations of nearly 700 individual hydrologic events. We found a positive response of NO 3 − and DOC to flow in forested streams, but watershed development led to a negative relationship between NO 3 − and discharge, and thus a decoupling of the overall NO 3 − and DOC responses to flow. On event and seasonal timescales, NO 3 − and DOC consistently displayed different behaviors. For example, in several streams, FDOM yield was greatest during summer storms while NO 3 − yield was greatest during winter storms. Most streams had generalizable storm NO 3 − and DOC responses, but differences in the timing of NO 3 − and DOC transport suggest different catchment sources. Further, certain events, including rain‐on‐snow and summer storms following dry antecedent conditions, yielded disproportionate NO 3 − responses. High‐frequency data allow for increased understanding of the processes controlling solute variability and will help reveal their responses to changing climatic regimes. [ABSTRACT FROM AUTHOR]

In this study, smart meter, water quality measuring sensors, and major operation skills are developed on the basis of real-time operation and management environment. Smart meter is designed by improvement of conventional water meter to secure accuracy and communication function. Finally, in order to obtain the information for the effective operation and management of water distribution systems, forecasting algorithm of abnormality water use is developed using Kalman filter and principal component analysis. Developed model is then implemented for real on-line measurement of the flow at the test district metered area (DMA) in southern city of Korea. [ABSTRACT FROM AUTHOR]

Abstract: A Cu2O-doped RuO2 sensing electrode (SE) for potentiometric detection of dissolved oxygen (DO) was prepared and its structure and electrochemical properties were analyzed by scanning electron microscopy (SEM), X-ray diffraction (XRD), X-ray photoelectron microscopy (XPS) and energy-dispersive spectroscopy (EDS) techniques. Cu2O-RuO2-SE displayed a linear DO response from 0.5 to 8.0ppm (log[O2], −4.73 to −3.59) within a temperature range of 9–30°C. The maximum sensitivity of −47.4mV/decade at 7.27 pH was achieved at 10mol% Cu2O. Experimental evaluation of the Cu2O-doped RuO2-SE demonstrated that the doping of RuO2 not only improves its structure but also enhances both sensor''s selectivity and antifouling properties. Selectivity measurements revealed that 10mol% Cu2O-doped RuO2-SE is insensitive to the presence of Na+, Mg2+, K+, Ca2+, NO3 −, PO4 2− and SO4 2− ions in the solution in the concentration range of 10−7–10−1 mol/l. [Copyright &y& Elsevier]

Accurate detection and identification of natural or intentional contamination events in a drinking water pipe is critical to drinking water supply security and health risk management. To use conventional water quality sensors for the purpose, we have explored a real-time event adaptive detection, identification and warning (READiw) methodology and examined it using pilot-scale pipe flow experiments of 11 chemical and biological contaminants each at three concentration levels. The tested contaminants include pesticide and herbicides (aldicarb, glyphosate and dicamba), alkaloids (nicotine and colchicine), E. coli in terrific broth, biological growth media (nutrient broth, terrific broth, tryptic soy broth), and inorganic chemical compounds (mercuric chloride and potassium ferricyanide). First, through adaptive transformation of the sensor outputs, contaminant signals were enhanced and background noise was reduced in time-series plots leading to detection and identification of all simulated contamination events. The improved sensor detection threshold was 0.1% of the background for pH and oxidation-reduction potential (ORP), 0.9% for free chlorine, 1.6% for total chlorine, and 0.9% for chloride. Second, the relative changes calculated from adaptively transformed residual chlorine measurements were quantitatively related to contaminant-chlorine reactivity in drinking water. We have shown that based on these kinetic and chemical differences, the tested contaminants were distinguishable in forensic discrimination diagrams made of adaptively transformed sensor measurements. [Copyright &y& Elsevier]

Abstract: Integrated water quality monitoring sensors were fabricated using micro-fabrication technologies, nanostructured Pt-doped RuO2 sensing electrode (SE) and an Ag/AgCl reference electrode (RE). The properties of RuO2–SE were evaluated for potentiometric detection of pH and dissolved oxygen (DO) at temperatures from 9 to 30°C. Pt-doped nanostructured RuO2–SEs were prepared on alumina sensor substrates. Their morphology was investigated by scanning electron microscopy (SEM), X-ray diffraction (XRD), electron dispersion spectroscopy (EDX), thermo-gravimetric/differential thermal analysis (TGA/DTA) and Fourier transform infrared spectroscopy (FTIR). It was found that morphology and surface compositions of nanostructured RuO2–SEs are sensitive to superoxide oxygen ions (O2 −). Sensors with Pt-doped RuO2–SE showed Nernstian response to pH from 2.0 to 13.0 and were also capable of measuring DO in the range of 0.6–8.0ppm. The measurement results show very good linearity and an excellent reproducibility was obtained during the 8 months of testing so far. Nernstian slope was 58mV/pH at a temperature of 23°C. Although RuO2–SEs have exhibited very good response time to pH changes, which was within 1–2s at a temperature of 23°C, as water temperature cooled down, the sensor response time increased and was about 8–10min at working temperature of 9°C. Influence of the hydrogen ion (H+) diffusion in RuO2 films on the output emf drift at pH measurements was also investigated. [Copyright &y& Elsevier]

Abstract: Nanostructured RuO2 sensing electrodes (SE) for water quality sensors were prepared on platinized alumina substrates. Their morphology was analysed by scanning electron microscopy (SEM), X-ray diffraction (XRD) and Fourier transform infrared spectroscopy (FTIR). The results show that morphology and surface compositions of RuO2-SEs are sensitive to superoxide oxygen ions (). Sensing properties of RuO2-SEs were investigated for potentiometric detection of dissolved oxygen and pH at temperatures from 9 to 30°C. Sensor was capable to measure dissolved oxygen from 0.6 to 8.0ppm and from 2 to 13 pH, respectively. The measurement results show very good linearity and an excellent reproducibility was obtained. Nernstian slope was 58mV/pH at 23°C for the RuO2-SEs. Although RuO2 film SEs have shown very good response time to the pH changes, which was within 1–2s at temperature 23°C, as water temperature cooled down, the sensor response time increased and was about 8–10min at a temperature of 9°C. [Copyright &y& Elsevier]

In the urban drainage sector, the problem of polluting discharges in sewers may act on the proper functioning of the sewer system, on the wastewater treatment plant reliability and on the receiving water body preservation. Therefore, the implementation of a chemical monitoring network is necessary to promptly detect and contain the event of contamination. Sensor location is usually an optimization exercise that is based on probabilistic or black-box methods and their efficiency is usually dependent on the initial assumption made on possible eligibility of nodes to become a monitoring point. It is a common practice to establish an initial non-informative assumption by considering all network nodes to have equal possibilities to allocate a sensor. In the present study, such a common approach is compared with different initial strategies to pre-screen eligible nodes as a function of topological and hydraulic information, and non-formal 'grey' information on the most probable locations of the contamination source. Such strategies were previously compared for conservative xenobiotic contaminations and now they are compared for a more difficult identification exercise: the detection of nonconservative immanent contaminants. The strategies are applied to a Bayesian optimization approach that demonstrated to be efficient in contamination source location. The case study is the literature network of the Storm Water Management Model (SWMM) manual, Example 8. The results show that the pre-screening and 'grey' information are able to reduce the computational effort needed to obtain the optimal solution or, with equal computational effort, to improve location efficiency. The nature of the contamination is highly relevant, affecting monitoring efficiency, sensor location and computational efforts to reach optimality. [ABSTRACT FROM AUTHOR]

• New analysis for clustering multivariate time series for hydrological events. • Time series clustering and 2-D hysteresis provide insight into storm event behavior. • The new clustering approach is applied to Vermont watershed data. • Synthetically generated C-Q data provide effective environmental model validation. Hydrological storm events are a primary driver for transporting water quality constituents such as suspended sediments and nutrients. Analyzing the concentration (C) of these water quality constituents in response to river discharge (Q), particularly when monitored at high temporal resolution during a hydrological event, helps to characterize the dynamics and flux of such constituents. A conventional approach to storm event analysis is to reduce C-Q time series to two-dimensional (2-D) hysteresis loops and analyze these 2-D patterns. While informative, this hysteresis loop approach has limitations because projecting the C-Q time series onto a 2-D plane obscures detail (e.g., temporal variation) associated with the C-Q relationships. In this paper, we address this limitation using a multivariate event time series (METS) clustering approach that is validated using synthetically generated event times series. The METS clustering is then applied to river discharge and suspended sediment data (acquired through turbidity-based monitoring) from six watersheds in the Lake Champlain Basin located in the northeastern United States, and results in identifying four common types of hydrological water quality events. Statistical analysis on the events partitioned by both methods (METS clustering and 2-D hysteresis classification) helped identify hydrometeorlogical features of common event types. In addition, the METS and hysteresis analysis were simultaneously applied to a regional Vermont dataset to highlight the complimentary nature of using them in tandem for hydrological event analysis. [ABSTRACT FROM AUTHOR]

Diatoms are among the dominant phytoplankters in marine and freshwater habitats, and important biomarkers of water quality, making their identification and classification one of the current challenges for environmental monitoring. To date, taxonomy of the species populating a water column is still conducted by marine biologists on the basis of their own experience. On the other hand, deep learning is recognized as the elective technique for solving image classification problems. However, a large amount of training data is usually needed, thus requiring the synthetic enlargement of the dataset through data augmentation. In the case of microalgae, the large variety of species that populate the marine environments makes it arduous to perform an exhaustive training that considers all the possible classes. However, commercial test slides containing one diatom element per class fixed in between two glasses are available on the market. These are usually prepared by expert diatomists for taxonomy purposes, thus constituting libraries of the populations that can be found in oceans. Here we show that such test slides are very useful for training accurate deep Convolutional Neural Networks (CNNs). We demonstrate the successful classification of diatoms based on a proper CNNs ensemble and a fully augmented dataset, i.e., creation starting from one single image per class available from a commercial glass slide containing 50 fixed species in a dry setting. This approach avoids the time-consuming steps of water sampling and labeling by skilled marine biologists. To accomplish this goal, we exploit the holographic imaging modality, which permits the accessing of a quantitative phase-contrast maps and a posteriori flexible refocusing due to its intrinsic 3D imaging capability. The network model is then validated by using holographic recordings of live diatoms imaged in water samples i.e., in their natural wet environmental condition. [ABSTRACT FROM AUTHOR]