Applying high-frequency surrogate measurements and a wavelet-ANN model to provide early warnings of rapid surface water quality anomalies

Crucial time of emergency monitoring for reliable numerical pollution source identification

The Pollution source identification (PSI) is an important issue on river water quality management especially for urban receiving water. Numerical inversion method is theoretically an effective PSI technique, which employs monitored downstream pollutant breakthrough curves to identify the pollution source. In practice, it is important to know how much monitoring data should be accumulated to provide PSI results with acceptable accuracy and uncertainty. However, no literature reports on this key point and it seriously handers the numerical PSI technology to mature practical applications. To seek a monitoring guideline for PSI, we conducted extensively numerical experiments for single-point source instantaneous release taking Bayesian-MCMC method as the baseline inversion technique. The crucial time (Tc) phenomenon was found during the data accumulation process for Bayesian source inversion. After Tc, estimated source parameters subsequent sustained low error levels and uncertainty convergence. Results shown the presence of Tc impacted by the number and location of monitoring sections, while monitoring frequency and data error do not. Under different river hydrodynamic conditions, relative crucial time (Λ) is determined by the river's Peclet number, and minimum effective Λ was controlled by dispersion coefficient (Dx). Analytic spatial structure of Λ(U, Dx) was uncovered and this relationship successfully explained by the information entropy theory. Based on these findings, a novel design method of PSI emergency monitoring network for preparedness plan and a practical framework of PSI for emergency response were established. These findings fill the important knowledge gap in PSI applications and the guidelines provide valuable references for river water quality management.

Innovative interpretable AI-guided water quality evaluation with risk adversarial analysis in river streams considering spatial-temporal effects

Water security remains a critical issue given the looming threats of industrial pollution, necessitating comprehensive assessments of water quality to address seasonal fluctuations and influential factors while formulating effective strategies for decision makers. This study introduces a novel approach for evaluating water quality within a complex riverine zone in South Korea: Han River that encompasses five river streams situated at each junction of North and South streams (including Gyeongan Stream) that ultimately leading towards Paldang Lake. By utilizing the monthly water characteristic data from the year 2013-2022 across 14 different locations, the significant seasonal trends and potential influences on water quality are identified. The water quality here is calculated with the proposed method of sub-index water quality index (s-WQI). A combinatorial prediction approach of s-WQI for each location is conducted through a collective of data preprocessing approaches including Hampel filtering and feature selection in prior to the machine learning predictions. In return, light gradient boosting (LGB) is the most accurate predictor by outperforming other prediction algorithms, especially through LGB-Pearson and LGB-Spearman combinations for North and South stream intersections, and LGB-Pearson for Paldang Lake. To further evaluate the robustness of this evaluation and extending the results to a foreseeable scenario, a seasonal based Monte-Carlo Simulation with 10,000 attempts targeting the water characteristic distributions obtained from each location considered are carried out to identify the risk bounds within. The results are further interpreted with SHAP analysis on identifying the contributions of each water characteristics towards the water quality through local and global spectrum. This research yields practical implications, offering tailored strategies for water quality enhancement and early warning systems. The integration of AI-based prediction and feature selection underscores the transformative potential of computational techniques in advancing data-driven water quality assessments, shaping the future of environmental science research.

Modeling benefits and tradeoffs of green infrastructure: Evaluating and extending parsimonious models for neighborhood stormwater planning

Green infrastructure is often proposed to complement conventional urban stormwater management systems that are stressed by extreme storms and expanding impervious surfaces. Established hydrological and hydraulic models inform stormwater engineering but are time- and data-intensive or aspatial, rendering them inadequate for rapid exploration of solutions. Simple spreadsheet models support quick site plan assessments but cannot adequately represent spatial interactions beyond a site. The present study builds on the Landscape Green Infrastructure Design (L-GrID) Model, a process-based spatial model that enables rapid development and exploration of green infrastructure scenarios to mitigate neighborhood flooding. We first explored how well L-GrID could replicate flooding reports in a neighborhood in Chicago, Illinois, USA, to evaluate its potential for green infrastructure planning. Although not meant for prediction, L-GrID was able to replicate the flooding reported and helped identify strategies for flood control. Once evaluated for this neighborhood, we extended the model to include water quality through the representation of dispersion and settling mechanisms for two pollutant surrogates-total nitrogen and total suspended solids. With the extended model, Landscape Green Infrastructure Design Model-Water Quality (L-GrID-WQ), we examined benefits, costs, and tradeoffs for different green infrastructure strategies. Bioswales were slightly more effective than other green infrastructure types in reducing flooding extent and downstream runoff and pollution, through increased infiltration and settling capacity. Permeable pavers followed in effectiveness and are suggested where spatial constraints may limit the installation of bioswales. Although green infrastructure supports both flooding and pollution control, small tradeoffs between these functions emerged across spatial layouts: strategies based on only curb-cuts better controlled pollution, while layouts that followed the path of water flow better controlled flooding. By illuminating such tradeoffs, L-GrID-WQ can support green infrastructure planning that prioritizes unique concerns in different areas of a landscape.

Identification of water pollution sources and analysis of pollution trigger conditions in Jiuqu River, Luxian County, China

Against the backdrop of ecological conservation and high-quality development in the Yangtze River Basin, there is an increasing demand for enhanced water pollution prevention and control in small watersheds. To delve deeper into the intricate relationship between pollutants and environmental features, as well as explore the key factors triggering pollution and their corresponding warning thresholds, this study was conducted along the Jiuqu River, a strategically managed unit in the upstream region of the Yangtze River, between 2022 and 2023. A total of seven monitoring sites were established, from which 161 valid water samples were collected. The k-nearest neighbors mutual information (KNN-MI) technique indicated that water temperature (WT) and relative humidity (RH) were the main environmental factors. The principal component analysis (PCA) of ten water quality parameters and three environmental factors unveiled the distinguishing characteristics of the primary pollution sources. Consequently, the pollution sources were categorized as treated wastewater > groundwater runoff > phytoplankton growth > abstersion wastewater > agricultural drainage. Furthermore, the regression decision tree (RDT) algorithm was used to explore the combined effects between pollutants and environmental factors, and to provide visual decision-making process and quantitative results for understanding the triggering mechanism of organic pollution in Jiuqu River. It conclusively identifies total phosphorus (TP) as the predominant triggering parameter with the threshold of 0.138 mg/L. The study is helpful to deal with potential water pollution problems preventatively and shows the interpretability and predictive performance of the RDT algorithm in water pollution prevention.

Physics-informed neural networks as surrogate models of hydrodynamic simulators

In response to growing concerns surrounding the relationship between climate change and escalating flood risk, there is an increasing urgency to develop precise and rapid flood prediction models. Although high-resolution flood simulations have made notable advancements, they remain computationally expensive, underscoring the need for efficient machine learning surrogate models. As a result of sparse empirical observation and expensive data collection, there is a growing need for the models to perform effectively in 'small-data' contexts, a characteristic typical of many scientific problems. This research combines the latest developments in surrogate modelling and physics-informed machine learning to propose a novel Physics-Informed Neural Network-based surrogate model for hydrodynamic simulators governed by Shallow Water Equations. The proposed method incorporates physics-based prior information into the neural network structure by encoding the conservation of mass into the model without relying on calculating continuous derivatives in the loss function. The method is demonstrated for a high-resolution inland flood simulation model and a large-scale regional tidal model. The proposed method outperforms the existing state-of-the-art data-driven approaches by up to 25 %. This research demonstrates the benefits and robustness of physics-informed approaches in surrogate modelling for flood and hydroclimatic modelling problems.

A coupled model to improve river water quality prediction towards addressing non-stationarity and data limitation

Accurate predictions of river water quality are vital for sustainable water management. However, even the powerful deep learning model, i.e., long short-term memory (LSTM), has difficulty in accurately predicting water quality dynamics owing to the high non-stationarity and data limitation in a changing environment. To wiggle out of quagmires, wavelet analysis (WA) and transfer learning (TL) techniques were introduced in this study to assist LSTM modeling, termed WA-LSTM-TL. Total phosphorus, total nitrogen, ammonia nitrogen, and permanganate index were predicted in a 4 h step within 49 water quality monitoring sites in a coastal province of China. We selected suitable source domains for each target domain using an innovatively proposed regionalization approach that included 20 attributes to improve the prediction efficiency of WA-LSTM-TL. The coupled WA-LSTM facilitated capturing non-stationary patterns of water quality dynamics and improved the performance by 53 % during testing phase compared to conventional LSTM. The WA-LSTM-TL, aided by the knowledge of source domain, obtained a 17 % higher performance compared to locally trained WA-LSTM, and such improvement was more impressive when local data was limited (+66 %). The benefit of TL-based modeling diminished as data quantity increased; however, it outperformed locally direct modeling regardless of whether target domain data was limited or sufficient. This study demonstrates the reasoning for coupling WA and TL techniques with LSTM models and provides a newly coupled modeling approach for improving short-term prediction of river water quality from the perspectives of non-stationarity and data limitation.

An integrated supervision framework to safeguard the urban river water quality supported by ICT and models

Models and information and communication technology (ICT) can assist in the effective supervision of urban receiving water bodies and drainage systems. Single model-based decision tools, e.g., water quality models and the pollution source identification (PSI) method, have been widely reported in this field. However, a systematic pathway for environmental decision support system (EDSS) construction by integrating advanced single techniques has rarely been reported, impeding engineering applications. This paper presents an integrated supervision framework (UrbanWQEWIS) involving monitoring-early warning-source identification-emergency disposal to safeguard the urban water quality, where the data, model, equipment and knowledge are smoothly and logically linked. The generic architecture, all-in-one equipment and three key model components are introduced. A pilot EDSS is developed and deployed in the Maozhou River, China, with the assistance of environmental Internet of Things (IoT) technology. These key model components are successfully validated via in situ monitoring data and dye tracing experiments. In particular, fluorescence fingerprint-based qualitative PSI and Bayesian-based quantitative PSI methods are effectively coupled, which can largely reduce system costs and enhance flexibility. The presented supervision framework delivers a state-of-the-art management tool in the digital water era. The proposed technical pathway of EDSS development provides a valuable reference for other regions.

Advances in Catchment Science, Hydrochemistry, and Aquatic Ecology Enabled by High-Frequency Water Quality Measurements

High-frequency water quality measurements in streams and rivers have expanded in scope and sophistication during the last two decades. Existing technology allows in situ automated measurements of water quality constituents, including both solutes and particulates, at unprecedented frequencies from seconds to subdaily sampling intervals. This detailed chemical information can be combined with measurements of hydrological and biogeochemical processes, bringing new insights into the sources, transport pathways, and transformation processes of solutes and particulates in complex catchments and along the aquatic continuum. Here, we summarize established and emerging high-frequency water quality technologies, outline key high-frequency hydrochemical data sets, and review scientific advances in key focus areas enabled by the rapid development of high-frequency water quality measurements in streams and rivers. Finally, we discuss future directions and challenges for using high-frequency water quality measurements to bridge scientific and management gaps by promoting a holistic understanding of freshwater systems and catchment status, health, and function.

Anomaly Detection in Biological Early Warning Systems Using Unsupervised Machine Learning

The use of bivalve mollusks as bioindicators in automated monitoring systems can provide real-time detection of emergency situations associated with the pollution of aquatic environments. The behavioral reactions of Unio pictorum (Linnaeus, 1758) were employed in the development of a comprehensive automated monitoring system for aquatic environments by the authors. The study used experimental data obtained by an automated system from the Chernaya River in the Sevastopol region of the Crimean Peninsula. Four traditional unsupervised machine learning techniques were implemented to detect emergency signals in the activity of bivalves: elliptic envelope, isolation forest (iForest), one-class support vector machine (SVM), and local outlier factor (LOF). The results showed that the use of the elliptic envelope, iForest, and LOF methods with proper hyperparameter tuning can detect anomalies in mollusk activity data without false alarms, with an F1 score of 1. A comparison of anomaly detection times revealed that the iForest method is the most efficient. These findings demonstrate the potential of using bivalve mollusks as bioindicators in automated monitoring systems for the early detection of pollution in aquatic environments.

Machine learning-based estimation of riverine nutrient concentrations and associated uncertainties caused by sampling frequencies

Accurate and sufficient water quality data is essential for watershed management and sustainability. Machine learning models have shown great potentials for estimating water quality with the development of online sensors. However, accurate estimation is challenging because of uncertainties related to models used and data input. In this study, random forest (RF), support vector machine (SVM), and back-propagation neural network (BPNN) models are developed with three sampling frequency datasets (i.e., 4-hourly, daily, and weekly) and five conventional indicators (i.e., water temperature (WT), hydrogen ion concentration (pH), electrical conductivity (EC), dissolved oxygen (DO), and turbidity (TUR)) as surrogates to individually estimate riverine total phosphorus (TP), total nitrogen (TN), and ammonia nitrogen (NH₄⁺-N) in a small-scale coastal watershed. The results show that the RF model outperforms the SVM and BPNN machine learning models in terms of estimative performance, which explains much of the variation in TP (79 ± 1.3%), TN (84 ± 0.9%), and NH₄⁺-N (75 ± 1.3%), when using the 4-hourly sampling frequency dataset. The higher sampling frequency would help the RF obtain a significantly better performance for the three nutrient estimation measures (4-hourly > daily > weekly) for R² and NSE values. WT, EC, and TUR were the three key input indicators for nutrient estimations in RF. Our study highlights the importance of high-frequency data as input to machine learning model development. The RF model is shown to be viable for riverine nutrient estimation in small-scale watersheds of important local water security.

Prediction of tempo-spatial patterns and exceedance probabilities of atmospheric corrosion of Q235 carbon steel across China

To reduce the losses caused by the atmospheric corrosion of carbon steels, it is important to establish a prediction model to determine the corrosion rate of carbon steels in natural environments. In this study, a prediction model of atmospheric corrosion of Q235 carbon steel (PMACC-Q235) in China was established by coupling the mean impact value algorithm and back propagation artificial neural network. Tempo-spatial patterns of corrosion rates in five long-exposure time categories across China were analyzed. Ten main factors affecting the atmospheric corrosion of Q235 were identified. The corrosion rates in a single year were similar (approximately 30 μm/a) and larger than those for 2 (25.30 μm/a) and 3 years (21.66 μm/a). The spatial corrosion rates in the northwestern areas were primarily lower than those in southeastern coastal areas. This could be influenced by climatic factors, such as temperature, humidity, and precipitation. All corrosion rates reached the C2 level (>1.3 μm/a), and there was some possibility that they reached higher corrosion levels. The largest probability for the C3 level in all periods was an average of 0.91, and that for the C4 level was 0.83. Spatially, higher probabilities were mainly located in the southern area, especially in Hainan, located in the south and surrounded by sea. Corrosion rates largely varied among climatic zones, and mean corrosion rates in the tropical monsoon climate zone were the largest (average of three periods 33.39 μm/a). SO₂ and soluble-dust fall had the largest impact on the variations in the corrosion rates among different climatic zones.

A Hybrid Model for Water Quality Prediction Based on an Artificial Neural Network, Wavelet Transform, and Long Short-Term Memory

Clean water is an indispensable essential resource on which humans and other living beings depend. Therefore, the establishment of a water quality prediction model to predict future water quality conditions has a significant social and economic value. In this study, a model based on an artificial neural network (ANN), discrete wavelet transform (DWT), and long short-term memory (LSTM) was constructed to predict the water quality of the Jinjiang River. Firstly, a multi-layer perceptron neural network was used to process the missing values based on the time series in the water quality dataset used in this research. Secondly, the Daubechies 5 (Db5) wavelet was used to divide the water quality data into low-frequency signals and high-frequency signals. Then, the signals were used as the input of LSTM, and LSTM was used for training, testing, and prediction. Finally, the prediction results were compared with the nonlinear auto regression (NAR) neural network model, the ANN-LSTM model, the ARIMA model, multi-layer perceptron neural networks, the LSTM model, and the CNN-LSTM model. The outcome indicated that the ANN-WT-LSTM model proposed in this study performed better than previous models in many evaluation indices. Therefore, the research methods of this study can provide technical support and practical reference for water quality monitoring and the management of the Jinjiang River and other basins.

Predicting polycyclic aromatic hydrocarbons in surface water by a multiscale feature extraction-based deep learning approach

Accurate and effective prediction of polycyclic aromatic hydrocarbons (PAHs) in surface water remains a substantial challenge due to the limited understanding of the dynamic processes. To assist integrated surface water management, a novel hybrid surface water PAH prediction model based on a two-stage decomposition approach and deep learning algorithm was proposed. Specifically, a two-stage decomposition technique consisting of complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and variational mode decomposition (VMD) was first introduced to decompose the data into several subsequences to extract the main fluctuations and trends of the PAH sequence. Subsequently, the deep learning algorithm long short-term memory (LSTM) was employed to explore the latent dynamic characteristics of each subsequence. Finally, the predicted values of the subsequences were integrated to obtain the final predicted results. An empirical study was conducted based on PAH data of eight major rivers in Saxony, Germany. The empirical results proved that the CEEMDAN-VMD-LSTM model outperformed other benchmark data-driven methods in predicting PAHs in surface water because it combined the advantages of two-stage decomposition and deep learning methods. The mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R²) of the model were 27.89, 37.92 and 0.85, respectively. The proposed hybrid method can achieve effective and accurate water quality prediction and is an effective tool for surface water management.

Reconstructing Missing and Anomalous Data Collected from High-Frequency In-Situ Sensors in Fresh Waters

In situ sensors that collect high-frequency data are used increasingly to monitor aquatic environments. These sensors are prone to technical errors, resulting in unrecorded observations and/or anomalous values that are subsequently removed and create gaps in time series data. We present a framework based on generalized additive and auto-regressive models to recover these missing data. To mimic sporadically missing (i) single observations and (ii) periods of contiguous observations, we randomly removed (i) point data and (ii) day- and week-long sequences of data from a two-year time series of nitrate concentration data collected from Arikaree River, USA, where synoptically collected water temperature, turbidity, conductance, elevation, and dissolved oxygen data were available. In 72% of cases with missing point data, predicted values were within the sensor precision interval of the original value, although predictive ability declined when sequences of missing data occurred. Precision also depended on the availability of other water quality covariates. When covariates were available, even a sudden, event-based peak in nitrate concentration was reconstructed well. By providing a promising method for accurate prediction of missing data, the utility and confidence in summary statistics and statistical trends will increase, thereby assisting the effective monitoring and management of fresh waters and other at-risk ecosystems.

Machine learning for anomaly detection in cyanobacterial fluorescence signals

Many drinking water utilities drawing from waters susceptible to harmful algal blooms (HABs) are implementing monitoring tools that can alert them to the onset of blooms. Some have invested in fluorescence-based online monitoring probes to measure phycocyanin, a pigment found in cyanobacteria, but it is not clear how to best use the data generated. Previous studies have focused on correlating phycocyanin fluorescence and cyanobacteria cell counts. However, not all utilities collect cell count data, making this method impossible to apply in some cases. Instead, this paper proposes a novel approach to determine when a utility needs to respond to a HAB based on machine learning by identifying anomalies in phycocyanin fluorescence data without the need for corresponding cell counts or biovolume. Four widespread and open source algorithms are evaluated on data collected at four buoys in Lake Erie from 2014 to 2019: local outlier factor (LOF), One-Class Support Vector Machine (SVM), elliptic envelope, and Isolation Forest (iForest). When trained on standardized historical data from 2014 to 2018 and tested on labelled 2019 data collected at each buoy, the One-Class SVM and elliptic envelope models both achieve a maximum average F1 score of 0.86 among the four datasets. Therefore, One-Class SVM and elliptic envelope are promising algorithms for detecting potential HABs using fluorescence data only.

Statistical comparison between SARIMA and ANN's performance for surface water quality time series prediction

The performance comparison studies of the autoregressive integrated moving average model (ARIMA) and the artificial neural network (ANN) were mostly carried out between the selected model structures through trial-and-error, strongly influenced by model structure uncertainty. This research aims to make up for this inadequacy. First, a surface water quality prediction case study including eight monitoring sites in China was introduced. Second, the ARIMA and ANN's performance was compared statistically between 6912 Seasonal ARIMA (SARIMA) and 110,592 feedforward ANN with different model structures, based on the mean square error (MSE) distributions depicted by boxplots. In a statistical view, the ANN models obtained a significantly lower median value and a more concentrated distribution of validation MSEs, which indicated lighter overfitting and better generalization ability. Furthermore, the optimal SARIMA models' performance is inferior to even the median of the ANN models in the case study. In contrast with the previous comparisons among selected models, the statistical comparison in this study shows lower uncertainty.

A novel dynamic multi-criteria ensemble selection mechanism applied to drinking water quality anomaly detection

The provision of clean and safe drinking water is a crucial task for water supply companies from all over the world. To this end, automatic anomaly detection plays a critical role in drinking water quality monitoring. Recent anomaly detection studies use techniques that focus on a single global objective. Yet, companies need solutions that better balance the trade-off between false positives (FPs), which lead to financial losses to water companies, and false negatives (FNs), which severely impact public health and damage the environment. This work proposes a novel dynamic multi-criteria ensemble selection mechanism to cope with both problems simultaneously: the non-dominated local class-specific accuracy (NLCA). Moreover, experiments rely on recent time series related classification metrics to assess the predictive performance. Results on data from a real-world water distribution system show that NLCA outperforms other ensemble learning and dynamic ensemble selection techniques by more than 15% in terms of time series related F₁ scores. As a conclusion, NLCA enables the development of stronger anomaly detection systems for drinking water quality monitoring. The proposed technique also offers a new perspective on dynamic ensemble selection, which can be applied to different classification tasks to balance conflicting criteria.

Detecting Technical Anomalies in High-Frequency Water-Quality Data Using Artificial Neural Networks

Anomaly detection (AD) in high-volume environmental data requires one to tackle a series of challenges associated with the typical low frequency of anomalous events, the broad-range of possible anomaly types, and local nonstationary environmental conditions, suggesting the need for flexible statistical methods that are able to cope with unbalanced high-volume data problems. Here, we aimed to detect anomalies caused by technical errors in water-quality (turbidity and conductivity) data collected by automated in situ sensors deployed in contrasting riverine and estuarine environments. We first applied a range of artificial neural networks that differed in both learning method and hyperparameter values, then calibrated models using a Bayesian multiobjective optimization procedure, and selected and evaluated the "best" model for each water-quality variable, environment, and anomaly type. We found that semi-supervised classification was better able to detect sudden spikes, sudden shifts, and small sudden spikes, whereas supervised classification had higher accuracy for predicting long-term anomalies associated with drifts and periods of otherwise unexplained high variability.

Water temperature forecasting based on modified artificial neural network methods: Two cases of the Yangtze River

Water temperature is a controlling indicator of river habitat since many physical, chemical and biological processes in rivers are temperature dependent. Highly precise and reliable predictions of water temperature are important for river ecological management. In this study, a hybrid model named BP_PSO3, based on the BPNN (back propagation neural network) optimized by the PSO (particle swarm optimization) algorithm, is proposed for water temperature prediction using air temperature (Ta), discharge (Q) and day of year (DOY) as input variables. The performance of the BP_PSO3 model was compared with that of the BP_PSO1 (with Ta as the input) and BP_PSO2 (with Ta and Q as the inputs) models to evaluate the importance of the inputs. In addition, a comparison among the BPNN, RBFNN (radial basis function neural network), WNN (wavelet neural network), GRNN (general regression neural network), ELMNN (Elman neural network), and BP_PSO-based models was carried out based on the MAE, RMSE, NSE and R². The eight artificial intelligence models were examined to predict the water temperature at the Cuntan and Datong stations in the Yangtze River. The results indicated that the hybrid BPNN-PSO3 model had a stronger ability to forecast water temperature under both normal and extreme drought conditions. Optimization by the PSO algorithm and the inclusion of Q and DOY could help capture river thermal dynamics more accurately. The findings of this study could provide scientific references for river water temperature forecasting and river ecosystem protection.

Estimation of Water Quality Parameters with High-Frequency Sensors Data in a Large and Deep Reservoir

High-frequency sensors can monitor water quality with high temporal resolution and without environmental influence. However, sensors for detecting key water quality parameters, such as total nitrogen(TN), total phosphorus(TP), and other water environmental parameters, are either not yet available or have attracted limited usage. By using a large number of high-frequency sensor and manual monitoring data, this study establishes regression equations that measure high-frequency sensor and key water quality parameters through multiple regression analysis. Results show that a high-frequency sensor can quickly and accurately estimate dynamic key water quality parameters by evaluating seven water quality parameters. An evaluation of the flux of four chemical parameters further proves that the multi-parameter sensor can efficiently estimate the key water quality parameters. However, due to the different optical properties and ecological bases of these parameters, the high-frequency sensor shows a better prediction performance for chemical parameters than for physical and biological parameters. Nevertheless, these results indicate that combining high-frequency sensor monitoring with regression equations can provide real-time and accurate water quality information that can meet the needs in water environment management and realize early warning functions.

A simulation-based approach to assess the power of trend detection in high- and low-frequency water quality records

To provide more precise understanding of water quality changes, continuous sampling is being used more in surface water quality monitoring networks. However, it remains unclear how much improvement continuous monitoring provides over spot sampling, in identifying water quality changes over time. This study aims (1) to assess our ability to detect trends using water quality data of both high and low frequencies and (2) to assess the value of using high-frequency data as a surrogate to help detect trends in other constituents. Statistical regression models were used to identify temporal trends and then to assess the trend detection power of high-frequency (15 min) and low-frequency (monthly) data for turbidity and electrical conductivity (EC) data collected across Victoria, Australia. In addition, we developed surrogate models to simulate five sediment and nutrients constituents from runoff, turbidity and EC. A simulation-based statistical approach was then used to the compare the power to detect trends between the low- and high-frequency water quality records. Results show that high-frequency sampling shows clear benefits in trend detection power for turbidity, EC, as well as simulated sediment and nutrients, especially over short data periods. For detecting a 1% annual trend with 5 years of data, up to 97% and 94% improvements on the trend detection probability are offered by high-frequency data compared with monthly data, for turbidity and EC, respectively. Our results highlight the benefits of upgrading monitoring networks with wider application of high-frequency sampling.

Concentration estimation of dissolved oxygen in Pearl River Basin using input variable selection and machine learning techniques

Dissolved oxygen (DO) concentration is an essential index for water environment assessment. Here, we present a modeling approach to estimate DO concentrations using input variable selection and data-driven models. Specifically, the input variable selection technique, the maximal information coefficient (MIC), was used to identify and screen the primary environmental factors driving variation in DO. The data-driven model, support vector regression (SVR), was then used to construct a robust model to estimate DO concentration. The approach was illustrated through a case study of the Pearl River Basin in China. We show that the MIC technique can effectively screen major local environmental factors affecting DO concentrations. MIC value tended to stabilize when the sample size >3000 and EC had the highest score with an MIC >0.3 at both of the stations. The variable-reduced datasets improved the performance of the SVR model by a reduction of 28.65% in RMSE, and increase of 22.16%, 56.27% in R², NSE, respectively, relative to complete candidate sets. The MIC-SVR model constructed at the tidal river network performed better than nontidal river network by a reduction of approximately 63.01% in RMSE, an increase of 62.36% in NSE, and R² >0.9. Overall, the proposed technique was able to handle nonlinearity among environmental factors and accurately estimate DO concentrations in tidal river network regions.

Explore the relationship between fish community and environmental factors by machine learning techniques

In the face of multiple habitat alterations originating from both natural and anthropogenic factors, the fast-changing environments pose significant challenges for maintaining ecosystem integrity. Machine learning is a powerful tool for modeling complex non-linear systems through exploratory data analysis. This study aims at exploring a machine learning-based approach to relate environmental factors with fish community for achieving sustainable riverine ecosystem management. A large number of datasets upon a wide variety of eco-environmental variables including river flow, water quality, and species composition were collected at various monitoring stations along the Xindian River of Taiwan during 2005 and 2012. Then the complicated relationship and scientific essences of these heterogonous datasets are extracted using machine learning techniques to have a more holistic consideration in searching a guiding reference useful for maintaining river-ecosystem integrity. We evaluate and select critical environmental variables by the analysis of variance (ANOVA) and the Gamma test (GT), and then we apply the adaptive network-based fuzzy inference system (ANFIS) for an estimation of fish bio-diversity using the Shannon Index (SI). The results show that the correlation between model estimation and the biodiversity index is higher than 0.75. The GT results demonstrate that biochemical oxygen demand (BOD), water temperature, total phosphorus (TP), and nitrate-nitrogen (NO₃-N) are important variables for biodiversity modeling. The ANFIS results further indicate lower BOD, higher TP, and larger habitat (flow regimes) would generally provide a more suitable environment for the survival of fish species. The proposed methodology not only possesses a robust estimation capacity but also can explore the impacts of environmental variables on fish biodiversity. This study also demonstrates that machine learning is a promising avenue toward sustainable environmental management in river-ecosystem integrity.

Efficiency evaluation of reverse osmosis desalination plant using hybridized multilayer perceptron with particle swarm optimization

The scarcity of freshwater causes the necessity for water delineation of brackish water. Reverse osmosis (RO) is one of the popular strategies characterized with lower cost and simple processing procedure compared to the other desalination techniques. The current research is conducted to investigate the efficiency the RO process based on one-week advance prediction of total dissolved solids (TDS) and permeate flow rate for Sistan and Bluchistan provinces located in Iran region. The water parameters including pH, feed pressure temperature, and conductivity are used to construct the prediction matrix. A newly hybrid data-intelligence (DI) model called multilayer perceptron hybridized with particle swarm optimization (MLP-PSO) is developed for the investigation. The potential of the proposed MLP-PSO model is validated against two predominate DI models including support vector machine (SVM) and M5Tree (M5T) models. The results evidenced the potential of the proposed MLP-PSO model over the SVM and M5T models in predicting the TDS and permeate flow rate. In addition, the proposed model attained lower uncertainty for the simulated data. Overall, the feasibility of the hybridized MLP-PSO achieved remarkable predictability for the RO process.

Tackling environmental challenges in pollution controls using artificial intelligence: A review

This review presents the developments in artificial intelligence technologies for environmental pollution controls. A number of AI approaches, which start with the reliable mapping of nonlinear behavior between inputs and outputs in chemical and biological processes in terms of prediction models to the emerging optimization and control algorithms that study the pollutants removal processes and intelligent control systems, have been developed for environmental clean-ups. The characteristics, advantages and limitations of AI methods, including single and hybrid AI methods, were overviewed. Hybrid AI methods exhibited synergistic effects, but with computational heaviness. The up-to-date review summarizes i) Various artificial neural networks employed in wastewater degradation process for the prediction of removal efficiency of pollutants and the search of optimizing experimental conditions; ii) Evaluation of fuzzy logic used for intelligent control of aerobic stage of wastewater treatment process; iii) AI-aided soft-sensors for precisely on-line/off-line estimation of hard-to-measure parameters in wastewater treatment plants; iv) Single and hybrid AI methods applied to estimate pollutants concentrations and design monitoring and early-warning systems for both aquatic and atmospheric environments; v) AI modelings of short-term, mid-term and long-term solid waste generations, and various ANNs for solid waste recycling and reduction. Finally, the future challenges of AI-based models employed in the environmental fields are discussed and proposed.

Development and evaluation of a real-time forecasting framework for daily water quality forecasts for Lake Chaohu to Lead time of six days

The socioeconomic benefits associated with informative water quality forecasts for large lakes are becoming increasingly evident. However, it remains an enormous challenge to produce forecasts of water quality variables that are accurate enough to meet public demand. In this study, we developed and evaluated a new forecast framework for real-time forecasting of daily dissolved oxygen (DO), ammonium nitrogen (NH), total phosphorus (TP) and total nitrogen (TN) concentrations at lead times from one to six days for Lake Chaohu, the fifth largest freshwater lake in China. The forecast framework is based on a 3-D hydrodynamic ecological model referred to as EcoLake. We used hydrological, meteorological and water quality data from multiple sources to generate initial conditions and forcing functions. Solar radiation and inflows from tributaries which are not readily available were calculated using forecasted cloud cover and rainfall. Forecast skill was evaluated based on 122 forecasts produced on different days in 2017 and for each of the 12 sampling sites. Results indicate that the skill of the forecast framework varies considerably across water quality variables, sampling sites, and lead times. Generally, the forecast framework is more skillful than the persistence forecasts, which use the most recent observations as forecasts. The TN forecasts tend to be the most skillful with a mean RMSE skill score of 28.5% averaged across the six lead times. The DO forecasts tend to have the lowest skill with an average value of 10.9%. Model sensitivity experiments further revealed that errors in the raw air temperature and wind speed forecasts have a noticeable impact on the overall skill of DO and NH forecasts. The forecast framework proposed here could be a useful operational forecasting tool to enhance the effectiveness of the drinking water supply and public health protection based on the water quality management of Lake Chaohu.

A data-driven model for real-time water quality prediction and early warning by an integration method

Due to increasingly serious deterioration of surface water quality, effective water quality prediction technique for real-time early warning is essential to guarantee the emergency response ability in advance for sustainable water management. In this study, an effective data-driven model for surface water quality prediction is developed to analyze the inherent water quality variation tendencies and provide real-time early warnings according to the historical observation data. The developed data-driven model is integrated by an improved genetic algorithm (IGA) for selecting optimal initial weight parameters of neural a network and a back-propagation neural network (BPNN) for adjusting appropriate connection architectures of neural network. First, improved genetic algorithm is used to optimize the reasonable initial weight parameters and prevent the developed model from selecting a local optimal result. Second, BPNN is applied to adjust appropriate connection architectures and identify the features of water quality variation. The developed model is then applied to forecast the surface water quality variations for real-time early warning in Ashi River, China, comparing with simple BPNN model. The prediction results demonstrate that the developed data-driven model can significantly improve the prediction performance both in prediction accuracy and reliability, and effectively provide real-time early warning for emergency response.

Integration Multi-Model to Evaluate the Impact of Surface Water Quality on City Sustainability: A Case from Maanshan City in China

Water pollution is a worldwide problem that needs to be solved urgently and has a significant impact on the efficiency of sustainable cities. The evaluation of water pollution is a Multiple Criteria Decision-Making (MCDM) problem and using a MCDM model can help control water pollution and protect human health. However, different evaluation methods may obtain different results. How to effectively coordinate them to obtain a consensus result is the main aim of this work. The purpose of this article is to develop an ensemble learning evaluation method based on the concept of water quality to help policy-makers better evaluate surface water quality. A valid application is conducted to illustrate the use of the model for the surface water quality evaluation problem, thus demonstrating the effectiveness and feasibility of the proposed model.

Determination of biochemical oxygen demand and dissolved oxygen for semi-arid river environment: application of soft computing models

Surface and ground water resources are highly sensitive aquatic systems to contaminants due to their accessibility to multiple-point and non-point sources of pollutions. Determination of water quality variables using mathematical models instead of laboratory experiments can have venerable significance in term of the environmental prospective. In this research, application of a new developed hybrid response surface method (HRSM) which is a modified model of the existing response surface model (RSM) is proposed for the first time to predict biochemical oxygen demand (BOD) and dissolved oxygen (DO) in Euphrates River, Iraq. The model was constructed using various physical and chemical variables including water temperature (T), turbidity, power of hydrogen (pH), electrical conductivity (EC), alkalinity, calcium (Ca), chemical oxygen demand (COD), sulfate (SO₄), total dissolved solids (TDS), and total suspended solids (TSS) as input attributes. The monthly water quality sampling data for the period 2004-2013 was considered for structuring the input-output pattern required for the development of the models. An advance analysis was conducted to comprehend the correlation between the predictors and predictand. The prediction performances of HRSM were compared with that of support vector regression (SVR) model which is one of the most predominate applied machine learning approaches of the state-of-the-art for water quality prediction. The results indicated a very optimistic modeling accuracy of the proposed HRSM model to predict BOD and DO. Furthermore, the results showed a robust alternative mathematical model for determining water quality particularly in a data scarce region like Iraq.

Novel hybrid linear stochastic with non-linear extreme learning machine methods for forecasting monthly rainfall a tropical climate

A novel hybrid approach is presented that can more accurately predict monthly rainfall in a tropical climate by integrating a linear stochastic model with a powerful non-linear extreme learning machine method. This new hybrid method was then evaluated by considering four general scenarios. In the first scenario, the modeling process is initiated without preprocessing input data as a base case. While in other three scenarios, the one-step and two-step procedures are utilized to make the model predictions more precise. The mentioned scenarios are based on a combination of stationarization techniques (i.e., differencing, seasonal and non-seasonal standardization and spectral analysis), and normality transforms (i.e., Box-Cox, John and Draper, Yeo and Johnson, Johnson, Box-Cox-Mod, log, log standard, and Manly). In scenario 2, which is a one-step scenario, the stationarization methods are employed as preprocessing approaches. In scenario 3 and 4, different combinations of normality transform, and stationarization methods are considered as preprocessing techniques. In total, 61 sub-scenarios are evaluated resulting 11013 models (10785 linear methods, 4 nonlinear models, and 224 hybrid models are evaluated). The uncertainty of the linear, nonlinear and hybrid models are examined by Monte Carlo technique. The best preprocessing technique is the utilization of Johnson normality transform and seasonal standardization (respectively) (R² = 0.99; RMSE = 0.6; MAE = 0.38; RMSRE = 0.1, MARE = 0.06, UI = 0.03 &UII = 0.05). The results of uncertainty analysis indicated the good performance of proposed technique (d-factor = 0.27; 95PPU = 83.57). Moreover, the results of the proposed methodology in this study were compared with an evolutionary hybrid of adaptive neuro fuzzy inference system (ANFIS) with firefly algorithm (ANFIS-FFA) demonstrating that the new hybrid methods outperformed ANFIS-FFA method.

Application of Least-Squares Support Vector Machines for Quantitative Evaluation of Known Contaminant in Water Distribution System Using Online Water Quality Parameters

In water-quality, early warning systems and qualitative detection of contaminants are always challenging. There are a number of parameters that need to be measured which are not entirely linearly related to pollutant concentrations. Besides the complex correlations between variable water parameters that need to be analyzed also impairs the accuracy of quantitative detection. In aspects of these problems, the application of least-squares support vector machines (LS-SVM) is used to evaluate the water contamination and various conventional water quality sensors quantitatively. The various contaminations may cause different correlative responses of sensors, and also the degree of response is related to the concentration of the injected contaminant. Therefore to enhance the reliability and accuracy of water contamination detection a new method is proposed. In this method, a new relative response parameter is introduced to calculate the differences between water quality parameters and their baselines. A variety of regression models has been examined, as result of its high performance, the regression model based on genetic algorithm (GA) is combined with LS-SVM. In this paper, the practical application of the proposed method is considered, controlled experiments are designed, and data is collected from the experimental setup. The measured data is applied to analyze the water contamination concentration. The evaluation of results validated that the LS-SVM model can adapt to the local nonlinear variations between water quality parameters and contamination concentration with the excellent generalization ability and accuracy. The validity of the proposed approach in concentration evaluation for potassium ferricyanide is proven to be more than 0.5 mg/L in water distribution systems.