Next Article in Journal
Cilantro Photosynthetic Parameters in Response to Different Flows of Nutrient Solutions Prepared with Brackish Waters Dominant in Na+, Cl, or Ca2+
Previous Article in Journal
Seasonal Variations of Hydraulic Exchange Between Surface Water and Groundwater in an Alluvial Plain Setting Using 222Rn
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Review

Water Quality Management in the Age of AI: Applications, Challenges, and Prospects

State Key Laboratory of Black Soils Conservation and Utilization, Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences, Changchun 130102, China
*
Authors to whom correspondence should be addressed.
Water 2025, 17(11), 1641; https://doi.org/10.3390/w17111641
Submission received: 17 April 2025 / Revised: 17 May 2025 / Accepted: 27 May 2025 / Published: 28 May 2025

Abstract

Artificial intelligence (AI) is transforming water environment management, creating new opportunities for improved monitoring, prediction, and intelligent regulation of water quality. This review highlights the transformative impact of AI, particularly through hybrid modeling frameworks that integrate AI with technologies like the Internet of Things (IoT), Remote Sensing (RS), and Unmanned Monitoring Platforms (UMP). These advances have significantly enhanced real-time monitoring accuracy, expanded the scope of data acquisition, and enabled comprehensive analysis through multisource data fusion. Coupling AI models with process-based models (PBM) has notably enhanced predictive capabilities for simulating water quality dynamics. Additionally, AI facilitates dynamic early-warning systems, precise pollutant source tracking, and data-driven decision-making. However, significant challenges remain, including data quality and accessibility, model interpretability, monitoring of hard-to-measure pollutants, and the lack of system integration and standardization. To address these bottlenecks, future research should focus on: (1) constructing high-quality, standardized open-access datasets; (2) developing explainable AI (XAI) models; (3) strengthening integration with digital twins and next-generation sensors; (4) improving the monitoring of trace and emerging pollutants; and (5) coupling AI with PBM by optimizing input data, internal mechanisms, and correcting model outputs through validation against observations. Overcoming these challenges will position AI as a central pillar in advancing smart water quality governance, safeguarding water security, and achieving sustainable development goals.

1. Introduction

The United Nations Sustainable Development Goal 6 (SDG 6) aims to ensure access to water and sanitation for all and promote their sustainable management. This goal emphasizes not only the availability of water resources but also the critical importance of water quality [1]. Amidst intensifying global urbanization, industrialization, and climate change, water pollution has emerged as an increasingly severe problem, placing immense pressure on water quality management [2,3]. Traditional water quality monitoring methods, which primarily rely on manual sampling and periodic laboratory analysis, suffer from several limitations. They are often inefficient, restricted in spatial coverage, and incapable of providing real-time responses to sudden pollution events. Consequently, these conventional approaches fail to meet the demands of modern water environment management for high-frequency, dynamic monitoring and rapid response capabilities [4]. Thus, the deep integration of Artificial Intelligence (AI) with the Internet of Things (IoT), Remote Sensing (RS), and big data analytics is increasingly recognized as a pivotal technological solution for enhancing the effectiveness of water quality monitoring and management [5].
AI is fundamentally reshaping the paradigm of water environment monitoring and management through its unparalleled capabilities [6]. It is projected that from 2024 to 2032, the size of the global water management market for AI will grow from 7.54 billion US dollars to 53.85 billion US dollars, at an annual growth rate of 27.85% [7]. Its primary advantage lies in the efficient processing of complex, high-dimensional water environment data. This enables near-real-time dynamic analysis, accurate identification of anomalous pollution events (e.g., sudden leaks or algal blooms), and forward-looking predictions of water quality trends based on historical data and environmental factors [8]. A detailed comparison of performance indicators for AI-driven and conventional water monitoring approaches is illustrated in Table 1. Ponnuru et al. [9] assessed the performance of four Machine Learning (ML) algorithms—Support Vector Machines (SVM), Decision Trees (DT), Artificial Neural Networks (ANN), and Random Forests (RF)—in water quality monitoring using IoT sensor data, RS imagery, and publicly available environmental datasets. The results showed that ANN achieved the highest accuracy (95.2%), followed by RF at 92.8%, SVM at 89.5%, and DT at 87.3%. The AI models effectively reduced the error rate by 30%, improved real-time monitoring efficiency by 40%, and significantly enhanced pollution detection capabilities, demonstrating the outstanding performance of AI in real-time pollution detection and data processing. However, a critical obstacle arises from the inherent “black-box” nature of many AI models, whose decision-making processes often lack transparency and intuitive explainability [10]. This lack of transparency poses a challenge when crucial decisions concerning water quality safety are required (such as identifying pollution sources, assessing health risks, or formulating remediation strategies), as decision-makers may hesitate to fully trust models with opaque reasoning processes. Furthermore, AI models, particularly Deep Learning (DL), are heavily dependent on large-scale, high-quality, and accurately labeled training datasets [5]. This significant data reliance implies that the predictive accuracy and stability of these models can be substantially compromised in situations lacking sufficient data or when encountering non-standard, unforeseen environmental conditions (like the emergence of rare pollutants or extreme weather events). Such limitations may result in misleading outcomes and pose risks to water environmental safety. Therefore, overcoming these bottlenecks—namely explainability and data dependency—is crucial for advancing AI from theoretical research to large-scale, high-reliability applications in the water quality domain.
Against this backdrop, this review systematically analyzes the current applications and recent advancements of AI in water quality monitoring, prediction, and management. The focus is placed on key technological areas such as real-time monitoring, predictive modeling, and management optimization. Specifically, it explores how AI enhances real-time monitoring accuracy, expands spatiotemporal coverage, improves prediction precision, and supports intelligent decision-making in water quality management. This article does not comprehensively cover all AI applications in water quality research, but instead focuses on current significant developments, highlighting successful experiences and limitations, and providing reference pathways for future research.

2. AI for Water Quality Monitoring

2.1. Application of AI in Real-Time Water Quality Monitoring

Traditional water quality monitoring methods, which rely on manual sampling and laboratory analysis, are inherently limited by low timeliness and insufficient spatiotemporal resolution [12]. With recent technological advancements, IoT-based sensors now facilitate continuous monitoring of various physical, chemical, and biological water quality parameters, transmitting data in real time to support more responsive and data-driven water management [17]. However, sensor data remains vulnerable to environmental disturbances, fluctuating hydrological conditions, and technical issues such as drift, malfunction, and data loss, particularly in complex environments (e.g., underwater or outdoor settings), which poses challenges to the accuracy and reliability of monitoring results [24,25]. Integrating AI—particularly ML algorithms—offers a promising solution to these challenges [26,27]. AI algorithms can intelligently process raw sensor data by performing real-time calibration, error correction, anomaly detection, and denoising, thereby significantly improving data quality and monitoring accuracy [8,28]. Additionally, AI’s powerful pattern recognition capabilities enable automatic diagnosis of sensor faults, self-calibration, and data compensation, thereby enhancing the stability and robustness of monitoring systems [29]. The deployment of AI models at sensor terminals (i.e., edge computing) allows for localized intelligent analysis of data, reducing transmission delays, improving response speed, and granting sensors the ability to adapt to environmental changes and optimize autonomously. This collaborative integration of AI, sensors, and IoT not only greatly improves the accuracy, real-time performance, and intelligence of water quality monitoring but also effectively compensates for the inherent limitations of sensors under complex conditions [16,30]. Table 2 summarizes current applications of AI algorithms in water quality monitoring, from traditional ML algorithms such as SVM and K-Means to DL models like Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM). AI technology is widely applied in key tasks such as water quality classification, pollutant concentration prediction, image-based pollution detection, and equipment fault diagnosis. These systems typically integrate IoT platforms with various sensors and combine RS or image data to enable low-cost, real-time remote data collection. Their significant advantages include high accuracy, real-time monitoring, low deployment costs, and strong environmental adaptability. The application fields cover various water environments such as drinking water, wastewater treatment, lakes, and aquaculture, significantly promoting the intelligent development of water quality monitoring and laying the foundation for building efficient water environment monitoring systems. Nonetheless, model interpretability and system-level integration remain critical challenges warranting further research.

2.2. AI Combined with RS and Unmanned Monitoring: Expanding New Dimensions in Water Quality Monitoring

While integrating AI with in-situ sensors has improved real-time monitoring accuracy, traditional monitoring stations still face limitations in spatial coverage and deployment flexibility. The adoption of RS technology and UMP offers broader spatial perspectives and greater operational flexibility in water quality monitoring.
RS technology is particularly valuable for providing large-scale, near-synchronous observations of surface water bodies, which is crucial for assessing the overall condition of vast water areas [39]. AI algorithms enable efficient analysis of RS imagery, allowing for precise extraction of surface water information, and enabling rapid identification of water quality issues such as algal blooms and pollution [40]. Chebud et al. [41] demonstrated that using ANN to process Landsat TM RS data allows for the monitoring of parameters such as chlorophyll-a (Chl-a), turbidity, and TP, providing a low-cost and efficient method for large-scale water body ecological assessments.
Unlike RS, which provides a macro perspective, the core advantage of UMP lies in their spatial deployment flexibility and the ability to perform close-range detection in specific areas. Unmanned systems, including Unmanned Aerial Systems (UAS), Unmanned Surface Vehicles (USV), and Unmanned Underwater Vehicles (UUV), significantly extend the spatial flexibility and accessibility of monitoring. These systems can access areas that are difficult to reach by traditional methods, such as remote water bodies, hazardous pollution zones, or disaster-affected areas [42]. Vasudevan and Baskaran [43] proposed an IoT-based USV system for real-time monitoring of parameters such as pH, temperature, and dissolved oxygen (DO) in remote or hard-to-reach water bodies. This system demonstrated advantages in low cost, low energy consumption, and high coverage, effectively overcoming the limitations of manual sampling and fixed monitoring stations in terms of coverage and maintenance. Ryu [42] developed an unmanned aerial platform that integrates monitoring, sampling, and visualization functions. Using the commercial drone DJI Matrice 600 Pro, equipped with 3D-printed sampling devices and sensors, this platform can monitor temperature, pH, conductivity, and DO in real time in remote or hazardous water areas. It also collects 1 L of water samples and transmits data to the cloud for visualization analysis via IoT technology. This system demonstrated the advantages of low cost, high mobility, and wide coverage, offering innovative support for environmental management.
In particular, UAS equipped with sensors for RS operations (i.e., drone RS) offers unique advantages for water quality monitoring [44]. Compared to satellite RS, drone-based RS can quickly obtain high spatial resolution image data at a lower cost and with greater flexibility, making it particularly suitable for detailed monitoring and investigation of small water bodies, shoreline areas, or specific pollution events. It effectively captures local water quality changes and supports high-frequency repeated observations.
Despite these technological advancements, current applications remain constrained by several key limitations: Most existing studies primarily focus on optically active variables, including Chl-a, colored dissolved organic matter (CDOM), TSS, and turbidity. However, many important water quality variables, such as pH, total nitrogen (TN), ammonia nitrogen (NH3-N), nitrate nitrogen (NO3-N), and dissolved phosphorus (DP), have not been well studied due to their weak optical properties and low signal-to-noise ratio. Additionally, data processing complexity: The massive data generated by RS imagery and unmanned devices places high demands on the computational capacity of AI algorithms. For example, satellite multispectral imagery often contains noise (e.g., cloud interference), necessitating complex preprocessing procedures. Equipment costs and maintenance: The procurement and maintenance costs of drones, USVs, and UUVs are relatively high [45], and they are easily affected by environmental factors (such as strong winds and water currents). For instance, the stability of USVs in harsh weather conditions still requires improvement. Spatial and depth limitations: Satellite RS has difficulty penetrating the water surface and is limited to monitoring the surface layer; although UUV can dive underwater, their coverage is limited, making large-scale deep monitoring challenging. Lack of technical standardization: The data formats from different RS platforms and unmanned devices are not unified, meaning that AI models need to be customized for specific data sources, which limits the universality of the systems.

2.3. Multi-Source Data Fusion and Intelligent Analysis

Although advanced technologies such as RS and UMP have greatly enhanced the frequency of water quality data collection and spatial coverage, relying on a single data source or simple data overlay is far from sufficient for transforming this abundant data into accurate assessments and effective decision-making. Given the complex and ever-changing water environment and the vast, heterogeneous data it generates, the combination of multi-source data fusion and intelligent analysis has become a key solution to overcoming bottlenecks and improving the intelligence level of water quality monitoring systems [46,47].
In the field of water quality monitoring, multi-source data fusion refers to the systematic integration of information from different channels to obtain a more comprehensive, reliable, and accurate understanding of water quality conditions than any single source alone [48,49]. These data sources can include:
High temporal resolution in-situ sensor data: Real-time or near-real-time measurements from fixed monitoring stations or mobile platforms (such as USV), including parameters like pH, DO, temperature, turbidity, etc.
Wide-area RS imagery data: Data from satellites or drones that provide spatial distribution information for surface water parameters such as Chl-a, suspended solids, water color index, and surface temperature.
High-precision laboratory analysis data: Chemical and biological tests conducted on collected water samples, such as for heavy metals, specific organic pollutants, and nutrients.
Other data: Including meteorological data (precipitation, temperature), hydrological data (flow rate, velocity), geographic information (land use, point-source pollution distribution), and socio-economic data.
By integrating these complementary data sources, the limitations of single data sources can be effectively overcome [50]. For example, RS can quickly locate potential problem areas, after which unmanned platforms can be deployed for close-range verification and sampling. This data can then be combined with continuous monitoring data from fixed stations and precise analysis results from laboratories, forming a spatially and temporally continuous, integrated monitoring network that combines both macro and micro perspectives [51].
The fused multi-source data typically exhibit characteristics of big data, including high volume (Volume), a variety of data types (Variety), fast velocity (Velocity), and low-value density (Value), making it challenging for traditional data processing methods to handle effectively [52]. In such cases, intelligent analysis technologies, represented by AI and ML, become particularly important. AI algorithms, especially DL models, excel at processing high-dimensional, nonlinear complex data and are capable of:
Automatic identification and extraction of hidden patterns: Discovering complex correlations and trends in the fused data that are difficult to detect through traditional methods.
Filling in data gaps: Estimating missing spatiotemporal data points by leveraging the relationships between available data.
Improving prediction accuracy: Building more robust water quality prediction models that provide early warnings for potential risks.
Supporting intelligent decision-making: Offering more reliable data support and decision recommendations for water environment management.
Jiang et al. [14] provide an excellent example of this approach. They successfully integrated data from various sources, including environmental, social, water quantity, and water quality data, using DL models to predict key pollution indicators in urban sewage networks (BOD5, COD, NH4+-N, TN, TP). The results showed that the model’s coefficient of determination (R2) ranged from 0.85 to 0.91, significantly outperforming traditional statistical methods. This clearly demonstrates the immense potential of AI-driven multi-source data fusion in extracting data value and improving water quality prediction accuracy.
Therefore, developing and applying multi-source data fusion technologies, combined with powerful AI-driven intelligent analysis capabilities, is a critical step in advancing modern water quality monitoring from “data collection” to “intelligent insights”. This not only greatly enhances the accuracy and reliability of monitoring results but also provides unprecedented decision support for water resource protection and environmental management.

3. The Application of AI in Water Quality Prediction

3.1. Water Quality Prediction Based on Historical and Real-Time Data

The development of water quality prediction models has evolved from traditional statistical regression to classic ML algorithms, and further to DL models. ML models, such as SVM and RF, have demonstrated a superior ability to predict key indicators like DO, chemical oxygen demand (COD), and TN compared to traditional statistical regressions (e.g., multiple linear regression (MLR), Autoregressive Integrated Moving Average ARIMA), particularly when dealing with complex nonlinear relationships between water quality parameters [53]. With advancements in time series data processing capabilities, water quality prediction models have gradually become more refined. He et al. [19] categorized these models into three patterns: the point-to-point (P2P) model, which uses synchronous data to predict the current state; the sequence-to-point (S2P) model, which uses historical sequences to predict future single points; and the sequence-to-sequence (S2S) model, which predicts future sequences based on historical data. The article discusses traditional ML methods (such as SVM, RF, multilayer perceptron neural networks (MLPNN), etc.) as well as DL methods (such as LSTM, CNN, etc.). The research indicates that DL models, especially LSTM and Gated Recurrent Units (GRU), have been widely applied in water quality prediction in recent years due to their advantages in handling time-series data. LSTM, an improved version of Recurrent Neural Networks (RNN), can effectively address the long-term dependency problem. In water quality prediction, LSTM can predict future water quality changes by inputting historical water quality data, particularly performing well with time-series water quality data [19,54]. AI technologies have also advanced the timeliness of water quality information acquisition. While traditional real-time monitoring relies on physical sensors, AI models can construct real-time simulation and prediction systems, providing rapid assessments of current and future water quality conditions. One example is Bang et al. [55], who developed a real-time water quality simulation toolbox using ML and application programming interfaces (APIs) to simulate and predict parameters such as DO, TN, and TP, presented in a graphical user interface (GUI). The near-real-time simulation and prediction results provided by this tool are crucial for timely responses to water quality changes and pollution event early warnings.
In addition to conventional water quality parameters, ML has gradually been applied to the study of pollutants that are difficult to measure online, such as emerging contaminants [56,57]. This includes the identification, environmental behavior prediction, toxicity analysis, and removal technology optimization of emerging pollutants like perfluoroalkyl substances, nanoparticles, antibiotic resistance genes, and microplastics [56]. By processing complex, high-dimensional, and unstructured data, ML enhances the accuracy and efficiency of pollutant detection, offering new approaches and methods for environmental pollution management.
To further improve prediction accuracy and address the complexity of real-world water systems, researchers have explored hybrid model strategies [58,59]. Rajaee et al. [60] reviewed the application of both single and hybrid AI-based models in river water quality prediction. The article evaluated several AI models, including ANN, Genetic Programming (GP), Fuzzy Logic (FL), and SVM, with particular emphasis on hybrid models such as Wavelet-Neural Networks (WANN) and Wavelet-Support Vector Regression (WSVR) for their advantages in enhancing prediction accuracy. The study found that ANN and WANN were the most commonly used models, especially excelling in predicting DO and suspended sediment load (SSL). Compared to traditional regression models (e.g., MLR, ARIMA), AI models are better at handling complex nonlinear relationships and noisy data.
Despite the powerful data mining capabilities demonstrated by AI in water quality prediction, its application still faces significant challenges. Purely data-driven models are highly sensitive to the quality and quantity of data, and they are constrained by issues such as missing data, outliers, and uneven spatial and temporal coverage [15]. The models themselves also exhibit uncertainties, such as the subjectivity involved in selecting indicators and assigning weights when constructing a comprehensive WQI, which may lead to predictive bias [21]. More critically, many existing AI models tend to focus on spatial points or time series predictions, making it difficult to ensure spatiotemporal consistency in the underlying physical processes. Furthermore, the “black-box” nature of the models limits their generalization ability and mechanistic interpretability, especially when confronted with long-term macro factors like climate change or sparse data scenarios, where their reliability is put to the test. On the other hand, traditional process-driven models (e.g., hydrodynamic-water quality models) are based on physical, chemical, and biological principles, and can simulate the complex two-dimensional or three-dimensional spatiotemporal dynamics of water bodies. These models have a solid mechanistic foundation and good interpretability. However, they tend to be structurally complex, computationally expensive, and require stringent parameterization and boundary condition settings, with the calibration process potentially introducing uncertainties. Therefore, coupling data-driven AI models with process-based physical models is an extremely promising research direction.

3.2. AI Coupled with Process-Based Models (PBM)

Traditional water quality models are primarily based on mathematical descriptions of physical, chemical, and biological processes. While they offer good mechanistic interpretability, they often face bottlenecks such as low computational efficiency and difficulty in parameter calibration when simulating complex, nonlinear water environment systems [61]. On the other hand, purely data-driven AI models, while possessing powerful fitting capabilities, lack physical constraints, potentially leading to predictions that do not align with natural laws [62]. The coupling of AI with process models combines the strengths of both approaches, enabling more efficient and accurate water quality prediction and management [63]. The core idea behind this “hybrid modeling” or “Process-Guided AI” (PGAI) is to leverage AI’s powerful data learning capabilities to capture complex patterns while incorporating the physical mechanisms of process models as constraints or prior knowledge. This approach not only significantly enhances prediction accuracy, robustness, and generalization ability, ensuring the physical consistency of results, but also effectively compensates for the shortcomings of purely data-driven models—such as the lack of mechanistic grounding—and the challenges associated with calibrating pure process models, which can be computationally expensive. It is particularly valuable in scenarios with sparse data [64].
As shown in Figure 1, effective approaches for coupling AI with process models include parameter optimization, hybrid modeling, and process-guided deep learning (PGDL) [65].
Parameter Optimization: AI optimization algorithms (e.g., genetic algorithms, particle swarm optimization) are used to automate and efficiently calibrate key parameters of traditional process models, replacing the tedious manual parameter adjustment process.
Hybrid Structural Modeling: Process models are combined with AI models (such as SVM) through serial, parallel, or feedback loops. The process model provides the physical framework or initial predictions, while the AI model corrects biases or handles complex nonlinear components.
PGDL: Physical process equations (e.g., differential equations) are directly embedded into the structure or loss function of DL models, allowing the model to simultaneously learn data patterns while adhering to physical laws during the training process.
This coupling strategy greatly enhances the ability to simulate the complex processes of water environments. For example, by combining DL with physically coupled hydrodynamic models, it becomes possible to more accurately characterize the spatiotemporal migration and transformation paths of pollutants and their concentration dynamics in complex river networks, significantly improving prediction accuracy and stability (as summarized in Figure 1) [66]. Similarly, integrating process models with data-driven models has been shown to effectively improve the prediction accuracy of emerging pollutants and allow for rapid assessments based on easily measurable water quality indicators, reducing computational costs [67]. Wang et al. [68] proposed a new framework that combines a process-based water quality model with a Modified Local Model (MLM) based on chaos theory for data assimilation, aimed at enhancing real-time forecasting of salinity and DO in Singapore’s coastal waters. Their approach significantly reduced the root-mean-square error (RMSE) in daily forecasts over 1–5 days, achieving RMSE values as low as 0.23 for DO and 0.36 for salinity. This demonstrates that the method effectively improves forecast accuracy for both short-term and long-term predictions. Further, Wang et al. [69] developed an integrated modeling framework at the watershed scale, combining hydrological, hydraulic, and water quality models, and improved water quality predictions using data assimilation techniques. In its application to the coastal waters of Singapore, this method significantly enhanced the prediction accuracy of Chl-a. Specifically, for a 2-day forecast, the RMSE of Chl-a decreased from 5.96 to 2.49, a reduction of 58%; for a 2-hour forecast, the RMSE decreased by 48%. These results indicate that data assimilation techniques substantially improved the model’s predictive capability, especially in real-time forecasting. This approach is not only applicable to long-term water quality planning and management but also to real-time water quality change predictions and event-driven water ecosystem changes. Additionally, the method was successfully applied to both coastal and freshwater systems in Singapore, providing a powerful tool for water quality management. A notable example of a successful AI-coupled PBM is the hybrid approach developed by Tong et al. [67], which combined a PBM with data-driven models (DDMs) to predict emerging contaminants (ECs) such as Bisphenol A (BPA) and N,N-diethyltoluamide (DEET) in a tropical reservoir. Their study developed 36 DDMs using ANN and RF, achieving high predictive accuracy. The relatively high R2, Nash-Sutcliffe Efficiency (NSE), low RMSE, and average relative deviation (ARD (%))—with R2~0.84, NSE~0.91, RMSE~0.12, ARD (%)~2.00—indicate that the models are highly capable of accurately predicting EC concentrations in the reservoir. Notably, this hybrid model reduced computational time by over 99% compared to standalone PBM, completing one-year simulations in approximately one minute, as opposed to 8 hours for PBM alone. This efficiency, coupled with the use of easily measurable water quality indicators (e.g., TN, TP, Chl-a), demonstrates the potential of hybrid models to provide rapid, cost-effective predictions for hard-to-measure pollutants, thereby supporting real-time water quality management in urban reservoirs. Likewise, coupling climate change models with AI models enables more reliable evaluations of the long-term potential impacts of future climate scenarios on water quality, providing scientific support for water resource adaptive management [70].
Despite the promising prospects, the widespread application of AI-coupled process models still faces challenges. The effective integration of heterogeneous data is a primary issue, as the two types of models often require different data types, resolutions, and formats. Moreover, the choice and optimization of coupling methods lack a systematic evaluation framework, requiring clear identification of the applicable boundaries and performance strengths and weaknesses of different approaches. Additionally, the computational burden imposed by complex coupled models is a major constraint.
Future research should focus on developing advanced data fusion technologies, establishing standardized coupling frameworks and evaluation systems, and using model simplification and high-performance computing (e.g., parallel computing, cloud computing) to improve computational efficiency. Through continuous technological innovation and interdisciplinary collaboration, AI-coupled process models are expected to become a more precise, efficient, and reliable core technological support in the fields of water quality prediction and smart water management, contributing key insights to ensure the health of water environments and the sustainable use of water resources.

4. The Application of AI in Smart Water Environment Management

AI is profoundly transforming water quality management, driving the shift from traditional passive responses to proactive prevention, precise prediction, and optimized decision-making. The integration of AI technology has significantly enhanced monitoring capabilities, operational efficiency, pollution control levels, and the scientific basis of overall management decisions [6].
The combination of AI and the IoT is a key pathway to achieving dynamic and efficient water management [17]. By integrating various water quality parameters collected in real-time by IoT sensors deployed in water bodies, AI algorithms—particularly ML and DL—are able to efficiently process these large-scale, heterogeneous data sets. This enables dynamic monitoring to track water quality changes in real time, trend forecasting based on existing data and historical patterns to predict future water quality conditions, and early warning systems to identify potential risks such as algal blooms or heavy metal contamination. These advancements strongly support the fundamental shift in management models from “reactive processing” to “proactive prevention”. At the same time, AI can optimize operations and maintenance by analyzing sensor data and equipment status, such as predicting potential failures (predictive maintenance) and optimizing maintenance schedules, thereby reducing equipment downtime and ensuring the stable operation of monitoring and treatment facilities [71]. Examples of successful integrated systems include the Korea Water Resources Corporation, which uses IoT, Geographic Information System (GIS), and AI for real-time monitoring, leak detection, and flood management, as well as the PUMAGUA project in Mexico, which leverages AI and IoT for water conservation and water quality improvement. These systems enable decision-makers to prioritize key areas or develop forward-looking management strategies based on real-time conditions and future forecasts [72,73].
AI has significantly enhanced the ability to identify and track pollution sources, a critical step in effective pollution management. By integrating and analyzing multi-source data from sensor networks, RS imagery, social media information, and official environmental reports, AI can more accurately pinpoint the origin of pollution events [51], simulate the migration and diffusion paths of pollutants in water bodies [63], and even quickly identify pollutant types by combining technologies like chemical spectroscopy [74]. The integration of AI with citizen science initiatives, such as crowdsourced data collection, offers a promising avenue for expanding spatial and temporal coverage of water quality observations [75]. Smartphone-enabled image recognition, low-cost sensors, and cloud platforms can empower local communities to participate in environmental monitoring, while AI algorithms can assist in validating, filtering, and analyzing these decentralized datasets to improve data quality and usability for decision-making [76]. This is particularly crucial for tracing the origins of complex ECs. Additionally, AI can help analyze the contributions of various pollution sources, such as industrial emissions and agricultural runoff, to the contamination of water bodies, providing a scientific basis for developing more targeted pollution control strategies [77].
The integration of AI with Decision Support Systems (DSS) has provided powerful intelligent support for water quality management. AI models, by learning from vast amounts of historical data and integrating real-time information, can not only predict water quality changes in real time but also offer data-driven optimization recommendations for water resource allocation and pollution control measures [20]. Furthermore, AI can assist in evaluating the potential effects of different management decisions through simulations, aiding in the selection of the best approach. This intelligent support makes the water quality management decision-making process more scientific, efficient, and accurate, laying a solid foundation for ensuring water resource safety and sustainable use. To visually summarize the overall process, Figure 2 illustrates a conceptual framework of AI-driven water quality management. This framework integrates data acquisition from IoT sensors, RS platforms, and citizen science; performs data fusion and intelligent analysis using ML and DL algorithms; and supports decision-making processes including risk forecasting, pollution source identification, and infrastructure optimization. The system operates as a dynamic feedback loop, enabling proactive and adaptive responses in real-time water governance.

5. Conclusions and Prospects

AI is rapidly emerging as a transformative force in modern water quality monitoring, prediction, and management, offering robust capabilities to address growing pressures on global water resources. As reviewed, the integration of AI with IoT, RS, and UMP has substantially advanced real-time water quality assessment, expanding spatiotemporal coverage and enabling comprehensive multi-source data analysis.
In predictive modeling, cutting-edge AI approaches, particularly DL and hybrid models combined with process-based approaches (i.e., PGAI), demonstrate superior performance in capturing complex nonlinear dynamics and predicting water quality trends compared to traditional methods. Moreover, AI is driving a paradigm shift in water quality governance by enabling a transition from reactive measures to proactive management. This includes dynamic early-warning systems, precise pollutant source identification, and optimized data-driven decision-making, significantly enhancing operational responsiveness and long-term strategic planning.
Despite these significant strides, realizing the full potential of AI in the water quality domain hinges on overcoming several persistent challenges (Figure 3). Critical obstacles include the scarcity of high-quality and standardized datasets essential for robust model training; the inherent lack of transparency in many complex AI models, which impedes trust and adoption for critical decisions; the ongoing difficulty in reliably monitoring hard-to-measure pollutants like emerging contaminants in real-time; and the complexities associated with integrating diverse data sources, platforms, and models seamlessly.
Addressing these bottlenecks is paramount for future progress. Research and development priorities should include: (1) constructing high-quality, standardized open-access datasets to ensure robust model training and enable effective benchmarking across studies; (2) developing explainable AI (XAI) models to enhance transparency, build stakeholder trust, and facilitate adoption for critical water management decisions; (3) strengthening integration with digital twins and next-generation sensors to create dynamic virtual replicas and provide richer data streams for more comprehensive system understanding and real-time control; (4) improving the monitoring of trace and emerging pollutants, leveraging AI to enhance detection from complex sensor signals and predict the occurrence of these hard-to-measure substances; and (5) coupling AI with PBM to synergistically combine data-driven insights with mechanistic understanding, thereby improving predictive accuracy and physical consistency by optimizing input data, internal mechanisms, and correcting model outputs through validation against observations.
By tackling these challenges through sustained interdisciplinary innovation and collaboration, AI is poised to play an increasingly pivotal role. Successfully navigating these hurdles will unlock AI’s capacity to significantly contribute to ensuring global water safety, achieving SDG 6, and ushering in an era of smarter, more precise, and adaptive water resource management.

Author Contributions

Conceptualization, H.J. and J.Z.; methodology, S.Z.; writing—original draft preparation, S.Z.; writing—review and editing, J.Z.; visualization, S.Z.; project administration, H.J. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National Natural Science Foundation of China (grant Nos. 4231101419 and 42471089), the Jilin Province International scientific and technological cooperation projects (grant No. 20240402026GH), the seventh batch Young Elite Scientists Sponsorship Program by Jilin Province (grant No. QT202330).

Data Availability Statement

No new data were created or analyzed in this study. Data sharing is not applicable to this article.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
AIArtificial intelligence
ANNArtificial Neural Networks
APIsapplication programming interfaces
ARDaverage relative deviation
ARIMAAutoregressive Integrated Moving Average
BOD55-day Biochemical Oxygen Demand
Boosting-IPW-PLSBoosting-Iterative Predictor Weighting-Partial Least Squares
BPABisphenol A
CDOMColored Dissolved Organic Matter
Chl-achlorophyll-a
CNNConvolutional Neural Networks
CODchemical oxygen demand
DDMsdata-driven models
DEETN,N-diethyltoluamide
DLDeep Learning
DOdissolved oxygen
DPdissolved phosphorus
DSSDecision Support Systems
DTDecision Trees
ECElectrical Conductivity
ECsemerging contaminants
FLFuzzy Logic
GISGeographic Information System
GPGenetic Programming
GRUGated Recurrent Unit
GUIgraphical user interface
IoTInternet of Things
K-MeansK-Means clustering
LSTMLong Short-Term Memory
MLMachine Learning
MLMModified Local Model
MLPNNmultilayer perceptron neural networks
MLRmultiple linear regression
NNNeural Network
NH3-NAmmonia Nitrogen
NH4+Ammonium nitrogen
NO3-NNitrate Nitrogen
NSENash-Sutcliffe Efficiency
ORPOxidation-Reduction Potential
P2Ppoint-to-point
PBMprocess-based models (or Process-Based Models)
PGAIProcess-Guided AI
PGDLprocess-guided deep learning
PSO-SVRparticle swarm optimization-support vector regression
R2coefficient of determination
RFRandom Forests
RNNRecurrent Neural Networks
RMSEroot-mean-square error
RSRemote Sensing
S2Psequence-to-point
S2Ssequence-to-sequence
SDG 6Sustainable Development Goal 6
SSLsuspended sediment load
SVMSupport Vector Machines
TDSTotal Dissolved Solids
TNTotal Nitrogen
TPTotal Phosphorus
TSSTotal Suspended Solids
UASUnmanned Aerial Systems
UMPUnmanned Monitoring Platforms
USVUnmanned Surface Vehicles
UUVUnmanned Underwater Vehicles
WANNWavelet-Neural Networks
WQIWater Quality Index
WSVRWavelet-Support Vector Regression
XAIexplainable AI

References

  1. Jones, E.R.; Bierkens, M.F.P.; Wanders, N.; Sutanudjaja, E.H.; van Beek, L.P.H.; van Vliet, M.T.H. Current Wastewater Treatment Targets Are Insufficient to Protect Surface Water Quality. Commun. Earth Environ. 2022, 3, 221. [Google Scholar] [CrossRef]
  2. Shi, X.; Mao, D.; Song, K.; Xiang, H.; Li, S.; Wang, Z. Effects of Landscape Changes on Water Quality: A Global Meta-Analysis. Water Res. 2024, 260, 121946. [Google Scholar] [CrossRef] [PubMed]
  3. Salerno, F.; Gaetano, V.; Gianni, T. Urbanization and Climate Change Impacts on Surface Water Quality: Enhancing the Resilience by Reducing Impervious Surfaces. Water Res. 2018, 144, 491–502. [Google Scholar] [CrossRef]
  4. Su, J.; Xu, W.; Lin, Z. Algorithm for Monitoring Water Quality Parameters in Optical Systems Based on Artificial Intelligence Data Mining. Sci. Rep. 2024, 14, 28142. [Google Scholar] [CrossRef]
  5. Essamlali, I.; Nhaila, H.; Khaili, M.E. Advances in Machine Learning and IoT for Water Quality Monitoring: A Comprehensive Review. Heliyon 2024, 10, e27920. [Google Scholar] [CrossRef] [PubMed]
  6. Sela, L.; Sowby, R.B.; Salomons, E.; Housh, M. Making Waves: The Potential of Generative AI in Water Utility Operations. Water Res. 2025, 272, 122935. [Google Scholar] [CrossRef]
  7. AI in Water Management Market Growth Rate, Industry Analysis with Key Companies 2025–2032. Available online: https://www.datamintelligence.com/research-report/ai-in-water-management-market (accessed on 13 May 2025).
  8. Olawade, D.B.; Wada, O.Z.; Ige, A.O.; Egbewole, B.I.; Olojo, A.; Oladapo, B.I. Artificial Intelligence in Environmental Monitoring: Advancements, Challenges, and Future Directions. Hyg. Environ. Health Adv. 2024, 12, 100114. [Google Scholar] [CrossRef]
  9. Ponnuru, A.; Madhuri, J.V.; Saravanan, S.; Vijayakumar, T.; Manimegalai, V.; Das, A. Data-Driven Approaches to Water Quality Monitoring: Leveraging AI, Machine Learning, and Management Strategies for Environmental Protection. J. Neonatal Surg. 2025, 14, 664–675. [Google Scholar] [CrossRef]
  10. Frincu, R.M. Artificial Intelligence in Water Quality Monitoring: A Review of Water Quality Assessment Applications. Water Qual. Res. J. 2024, 60, 164–176. [Google Scholar] [CrossRef]
  11. Bagheri, M.; Farshforoush, N.; Bagheri, K.; Shemirani, A.I. Applications of Artificial Intelligence Technologies in Water Environments: From Basic Techniques to Novel Tiny Machine Learning Systems. Process Saf. Environ. Prot. 2023, 180, 10–22. [Google Scholar] [CrossRef]
  12. Gholizadeh, M.H.; Melesse, A.M.; Reddi, L. A Comprehensive Review on Water Quality Parameters Estimation Using Remote Sensing Techniques. Sensors 2016, 16, 1298. [Google Scholar] [CrossRef] [PubMed]
  13. Nallakaruppan, M.K.; Gangadevi, E.; Shri, M.L.; Balusamy, B.; Bhattacharya, S.; Selvarajan, S. Reliable Water Quality Prediction and Parametric Analysis Using Explainable AI Models. Sci. Rep. 2024, 14, 7520. [Google Scholar] [CrossRef] [PubMed]
  14. Jiang, Y.; Li, C.; Sun, L.; Guo, D.; Zhang, Y.; Wang, W. A Deep Learning Algorithm for Multi-Source Data Fusion to Predict Water Quality of Urban Sewer Networks. J. Clean. Prod. 2021, 318, 128533. [Google Scholar] [CrossRef]
  15. Wai, K.P.; Chia, M.Y.; Koo, C.H.; Huang, Y.F.; Chong, W.C. Applications of Deep Learning in Water Quality Management: A State-of-the-Art Review. J. Hydrol. 2022, 613, 128332. [Google Scholar] [CrossRef]
  16. Durgun, Y. Real-Time Water Quality Monitoring Using AI-Enabled Sensors: Detection of Contaminants and UV Disinfection Analysis in Smart Urban Water Systems. J. King Saud Univ.-Sci. 2024, 36, 103409. [Google Scholar] [CrossRef]
  17. Miller, T.; Durlik, I.; Kostecka, E.; Kozlovska, P.; Łobodzińska, A.; Sokołowska, S.; Nowy, A. Integrating Artificial Intelligence Agents with the Internet of Things for Enhanced Environmental Monitoring: Applications in Water Quality and Climate Data. Electronics 2025, 14, 696. [Google Scholar] [CrossRef]
  18. Kamyab, H.; Khademi, T.; Chelliapan, S.; SaberiKamarposhti, M.; Rezania, S.; Yusuf, M.; Farajnezhad, M.; Abbas, M.; Hun Jeon, B.; Ahn, Y. The Latest Innovative Avenues for the Utilization of Artificial Intelligence and Big Data Analytics in Water Resource Management. Results Eng. 2023, 20, 101566. [Google Scholar] [CrossRef]
  19. He, M.; Qian, Q.; Liu, X.; Zhang, J.; Curry, J. Recent Progress on Surface Water Quality Models Utilizing Machine Learning Techniques. Water 2024, 16, 3616. [Google Scholar] [CrossRef]
  20. Soori, M.; Jough, F.K.G.; Dastres, R.; Arezoo, B. AI-Based Decision Support Systems in Industry 4.0, A Review. J. Econ. Technol. 2024, in press. [Google Scholar] [CrossRef]
  21. Yan, X.; Zhang, T.; Du, W.; Meng, Q.; Xu, X.; Zhao, X. A Comprehensive Review of Machine Learning for Water Quality Prediction over the Past Five Years. J. Mar. Sci. Eng. 2024, 12, 159. [Google Scholar] [CrossRef]
  22. Lu, H.-Y.; Cheng, C.-Y.; Cheng, S.-C.; Cheng, Y.-H.; Lo, W.-C.; Jiang, W.-L.; Nan, F.-H.; Chang, S.-H.; Ubina, N.A. A Low-Cost AI Buoy System for Monitoring Water Quality at Offshore Aquaculture Cages. Sensors 2022, 22, 4078. [Google Scholar] [CrossRef] [PubMed]
  23. Sanya, W.M.; Alawi, M.A.; Eugenio, I. Design and Development of Smart Water Quality Monitoring System Using IoT. Int. J. Adv. Sci. Res. Eng. (IJASRE) 2022, 8, 1–13. [Google Scholar] [CrossRef]
  24. Shakunt, P.S. Detection of Faults Based on Machine Learning Schemes in Wireless Sensor Networks. In Proceedings of the Signal Processing, Telecommunication and Embedded Systems with AI and ML Applications; Bhateja, V., Chakravarthy, V.V.S.S.S., Anguera, J., Ghosh, A., Flores Fuentes, W., Eds.; Springer Nature: Singapore, 2025; pp. 127–137. [Google Scholar]
  25. Shao, X.; Cai, B.; Zou, Z.; Shao, H.; Yang, C.; Liu, Y. Artificial Intelligence Enhanced Fault Prediction with Industrial Incomplete Information. Mech. Syst. Signal Process. 2025, 224, 112063. [Google Scholar] [CrossRef]
  26. Das, S.; Khondakar, K.R.; Mazumdar, H.; Kaushik, A.; Mishra, Y.K. AI and IoT: Supported Sixth Generation Sensing for Water Quality Assessment to Empower Sustainable Ecosystems. ACS EST Water 2025, 5, 490–510. [Google Scholar] [CrossRef]
  27. Forhad, H.M.; Uddin, M.D.R.; Chakrovorty, R.S.; Ruhul, A.M.; Faruk, H.M.; Kamruzzaman, S.; Sharmin, N.; Jamal, A.S.I.M.; Haque, Md.M.-U.; Morshed, A.M. IoT Based Real-Time Water Quality Monitoring System in Water Treatment Plants (WTPs). Heliyon 2024, 10, e40746. [Google Scholar] [CrossRef]
  28. Sharma, R.; Satapathy, A.; Srivastava, V.; Saxena, R. Revolutionizing Water Quality Management: The Impact of Machine Learning and Artificial Intelligence. In Computational Automation for Water Security; Dubey, A.K., Srivastav, A.L., Kumar, A., Garcia Marquez, F.P., Giannakoudakis, D.A., Eds.; Elsevier: Amsterdam, The Netherlands, 2025; pp. 27–42. [Google Scholar]
  29. Desikan, J.; Singh, S.K.; Jayanthiladevi, A.; Bhushan, S.; Rishiwal, V.; Kumar, M. Hybrid Machine Learning-Based Fault-Tolerant Sensor Data Fusion and Anomaly Detection for Fire Risk Mitigation in IIoT Environment. Sensors 2025, 25, 2146. [Google Scholar] [CrossRef]
  30. Lakshmikantha, V.; Hiriyannagowda, A.; Manjunath, A.; Patted, A.; Basavaiah, J.; Anthony, A.A. IoT Based Smart Water Quality Monitoring System. Glob. Transit. Proc. 2021, 2, 181–186. [Google Scholar] [CrossRef]
  31. Adeleke, I.A.; Nwulu, N.I.; Ogbolumani, O.A. A Hybrid Machine Learning and Embedded IoT-Based Water Quality Monitoring System. Internet Things 2023, 22, 100774. [Google Scholar] [CrossRef]
  32. Gupta, S.; Kohli, M.; Kumar, R.; Bandral, S. IoT Based Underwater Robot for Water Quality Monitoring. IOP Conf. Ser. Mater. Sci. Eng. 2021, 1033, 012013. [Google Scholar] [CrossRef]
  33. Wang, X.; Zhang, F.; Ding, J. Evaluation of Water Quality Based on a Machine Learning Algorithm and Water Quality Index for the Ebinur Lake Watershed, China. Sci. Rep. 2017, 7, 12858. [Google Scholar] [CrossRef]
  34. Gunda, N.S.K.; Gautam, S.H.; Mitra, S.K. Editors’ Choice—Artificial Intelligence Based Mobile Application for Water Quality Monitoring. J. Electrochem. Soc. 2019, 166, B3031. [Google Scholar] [CrossRef]
  35. Wu, Y.; Zhang, X.; Xiao, Y.; Feng, J. Attention Neural Network for Water Image Classification under IoT Environment. Appl. Sci. 2020, 10, 909. [Google Scholar] [CrossRef]
  36. Qin, X.; Gao, F.; Chen, G. Wastewater Quality Monitoring System Using Sensor Fusion and Machine Learning Techniques. Water Res. 2012, 46, 1133–1144. [Google Scholar] [CrossRef] [PubMed]
  37. Cecconi, F.; Rosso, D. Soft Sensing for On-Line Fault Detection of Ammonium Sensors in Water Resource Recovery Facilities. Environ. Sci. Technol. 2021, 55, 10067–10076. [Google Scholar] [CrossRef]
  38. Foschi, J.; Turolla, A.; Antonelli, M. Soft Sensor Predictor of E. Coli Concentration Based on Conventional Monitoring Parameters for Wastewater Disinfection Control. Water Res. 2021, 191, 116806. [Google Scholar] [CrossRef]
  39. Anand, V.; Oinam, B.; Wieprecht, S. Machine Learning Approach for Water Quality Predictions Based on Multispectral Satellite Imageries. Ecol. Inform. 2024, 84, 102868. [Google Scholar] [CrossRef]
  40. Rahat, S.H.; Steissberg, T.; Chang, W.; Chen, X.; Mandavya, G.; Tracy, J.; Wasti, A.; Atreya, G.; Saki, S.; Bhuiyan, M.A.E.; et al. Remote Sensing-Enabled Machine Learning for River Water Quality Modeling under Multidimensional Uncertainty. Sci. Total Environ. 2023, 898, 165504. [Google Scholar] [CrossRef]
  41. Chebud, Y.; Naja, G.M.; Rivero, R.G.; Melesse, A.M. Water Quality Monitoring Using Remote Sensing and an Artificial Neural Network. Water Air Soil Pollut. 2012, 223, 4875–4887. [Google Scholar] [CrossRef]
  42. Ryu, J.H. UAS-Based Real-Time Water Quality Monitoring, Sampling, and Visualization Platform (UASWQP). HardwareX 2022, 11, e00277. [Google Scholar] [CrossRef]
  43. Vasudevan, S.K.; Baskaran, B. An Improved Real-Time Water Quality Monitoring Embedded System with IoT on Unmanned Surface Vehicle. Ecol. Inform. 2021, 65, 101421. [Google Scholar] [CrossRef]
  44. Zhou, T.; Ma, S.; Liu, T.; Yao, S.; Li, S.; Gao, Y. Integrating UAV-Based Multispectral Data and Transfer Learning for Soil Moisture Prediction in the Black Soil Region of Northeast China. Agronomy 2025, 15, 759. [Google Scholar] [CrossRef]
  45. Ihuoma, S.O.; Madramootoo, C.A.; Kalacska, M. Integration of Satellite Imagery and in Situ Soil Moisture Data for Estimating Irrigation Water Requirements. Int. J. Appl. Earth Obs. Geoinf. 2021, 102, 102396. [Google Scholar] [CrossRef]
  46. Gunia, M.; Laine, M.; Malve, O.; Kallio, K.; Kervinen, M.; Anttila, S.; Kotamäki, N.; Siivola, E.; Kettunen, J.; Kauranne, T. Data Fusion System for Monitoring Water Quality: Application to Chlorophyll-a in Baltic Sea Coast. Environ. Model. Softw. 2022, 155, 105465. [Google Scholar] [CrossRef]
  47. Bigdeli, B.; Pahlavani, P.; Amirkolaee, H.A. An Ensemble Deep Learning Method as Data Fusion System for Remote Sensing Multisensor Classification. Appl. Soft Comput. 2021, 110, 107563. [Google Scholar] [CrossRef]
  48. Na, M.; Liu, X.; Tong, Z.; Sudu, B.; Zhang, J.; Wang, R. Analysis of Water Quality Influencing Factors under Multi-Source Data Fusion Based on PLS-SEM Model: An Example of East-Liao River in China. Sci. Total Environ. 2024, 907, 168126. [Google Scholar] [CrossRef] [PubMed]
  49. Priyanka, E.B.; Thangavel, S.; Mohanasundaram, R.; Anand, R. Solar Powered Integrated Multi Sensors to Monitor Inland Lake Water Quality Using Statistical Data Fusion Technique with Kalman Filter. Sci. Rep. 2024, 14, 25202. [Google Scholar] [CrossRef]
  50. Liang, Y.; Ding, F.; Liu, L.; Yin, F.; Hao, M.; Kang, T.; Zhao, C.; Wang, Z.; Jiang, D. Monitoring Water Quality Parameters in Urban Rivers Using Multi-Source Data and Machine Learning Approach. J. Hydrol. 2025, 648, 132394. [Google Scholar] [CrossRef]
  51. Duan, Q.; Zhang, Q.; Quan, X.; Zhang, H.; Huang, L. Innovations of Water Pollution Traceability Technology with Artificial Intelligence. Earth Crit. Zone 2024, 1, 100009. [Google Scholar] [CrossRef]
  52. Di Curzio, D.; Castrignanò, A.; Fountas, S.; Romić, M.; Viscarra Rossel, R.A. Multi-Source Data Fusion of Big Spatial-Temporal Data in Soil, Geo-Engineering and Environmental Studies. Sci. Total Environ. 2021, 788, 147842. [Google Scholar] [CrossRef]
  53. Zhu, M.; Wang, J.; Yang, X.; Zhang, Y.; Zhang, L.; Ren, H.; Wu, B.; Ye, L. A Review of the Application of Machine Learning in Water Quality Evaluation. Eco-Environ. Health 2022, 1, 107–116. [Google Scholar] [CrossRef]
  54. Rana, R.; Kalia, A.; Boora, A.; Alfaisal, F.M.; Alharbi, R.S.; Berwal, P.; Alam, S.; Khan, M.A.; Qamar, O. Artificial Intelligence for Surface Water Quality Evaluation, Monitoring and Assessment. Water 2023, 15, 3919. [Google Scholar] [CrossRef]
  55. Bang, G.-H.; Gwon, N.-H.; Cho, M.-J.; Park, J.-Y.; Baek, S.-S. Developing a Real-Time Water Quality Simulation Toolbox Using Machine Learning and Application Programming Interface. J. Environ. Manag. 2025, 377, 124719. [Google Scholar] [CrossRef] [PubMed]
  56. Lei, L.; Pang, R.; Han, Z.; Wu, D.; Xie, B.; Su, Y. Current Applications and Future Impact of Machine Learning in Emerging Contaminants: A Review. Crit. Rev. Environ. Sci. Technol. 2023, 53, 1817–1835. [Google Scholar] [CrossRef]
  57. Fan, J.; Huang, G.; Chi, M.; Shi, Y.; Jiang, J.; Feng, C.; Yan, Z.; Xu, Z. Prediction of Chemical Reproductive Toxicity to Aquatic Species Using a Machine Learning Model: An Application in an Ecological Risk Assessment of the Yangtze River, China. Sci. Total Environ. 2021, 796, 148901. [Google Scholar] [CrossRef] [PubMed]
  58. Yang, L.; Driscol, J.; Sarigai, S.; Wu, Q.; Lippitt, C.D.; Morgan, M. Towards Synoptic Water Monitoring Systems: A Review of AI Methods for Automating Water Body Detection and Water Quality Monitoring Using Remote Sensing. Sensors 2022, 22, 2416. [Google Scholar] [CrossRef]
  59. Wang, K.; Liu, L.; Ben, X.; Jin, D.; Zhu, Y.; Wang, F. Hybrid Deep Learning Based Prediction for Water Quality of Plain Watershed. Environ. Res. 2024, 262, 119911. [Google Scholar] [CrossRef]
  60. Rajaee, T.; Khani, S.; Ravansalar, M. Artificial Intelligence-Based Single and Hybrid Models for Prediction of Water Quality in Rivers: A Review. Chemom. Intell. Lab. Syst. 2020, 200, 103978. [Google Scholar] [CrossRef]
  61. Varadharajan, C.; Appling, A.P.; Arora, B.; Christianson, D.S.; Hendrix, V.C.; Kumar, V.; Lima, A.R.; Müller, J.; Oliver, S.; Ombadi, M.; et al. Can Machine Learning Accelerate Process Understanding and Decision-Relevant Predictions of River Water Quality? Hydrol. Process. 2022, 36, e14565. [Google Scholar] [CrossRef]
  62. Huang, S.; Xia, J.; Wang, Y.; Wang, G.; She, D.; Lei, J. Pollution Loads in the Middle-Lower Yangtze River by Coupling Water Quality Models with Machine Learning. Water Res. 2024, 263, 122191. [Google Scholar] [CrossRef]
  63. Huan, J.; Fan, Y.; Xu, X.; Zhou, L.; Zhang, H.; Zhang, C.; Hu, Q.; Cai, W.; Ju, H.; Gu, S. Deep Learning Model Based on Coupled SWAT and Interpretable Methods for Water Quality Prediction under the Influence of Non-Point Source Pollution. Comput. Electron. Agric. 2025, 231, 109985. [Google Scholar] [CrossRef]
  64. Read, J.S.; Jia, X.; Willard, J.; Appling, A.P.; Zwart, J.A.; Oliver, S.K.; Karpatne, A.; Hansen, G.J.A.; Hanson, P.C.; Watkins, W.; et al. Process-Guided Deep Learning Predictions of Lake Water Temperature. Water Resour. Res. 2019, 55, 9173–9190. [Google Scholar] [CrossRef]
  65. Bikmukhametov, T.; Jäschke, J. Combining Machine Learning and Process Engineering Physics towards Enhanced Accuracy and Explainability of Data-Driven Models. Comput. Chem. Eng. 2020, 138, 106834. [Google Scholar] [CrossRef]
  66. Jia, L.; Yen, N.; Pei, Y. A Physics-Coupled Deep Learning Framework for Hydrodynamic Diffusion Modeling in Watershed Systems: Integrating Spatiotemporal Networks and Environmental Constraints. IEEE Access 2025, 13, 34985–35003. [Google Scholar] [CrossRef]
  67. Tong, X.; You, L.; Zhang, J.; He, Y.; Gin, K.Y.-H. Advancing Prediction of Emerging Contaminants in a Tropical Reservoir with General Water Quality Indicators Based on a Hybrid Process and Data-Driven Approach. J. Hazard. Mater. 2022, 430, 128492. [Google Scholar] [CrossRef]
  68. Wang, X.; Zhang, J.; Babovic, V. Improving Real-Time Forecasting of Water Quality Indicators with Combination of Process-Based Models and Data Assimilation Technique. Ecol. Indic. 2016, 66, 428–439. [Google Scholar] [CrossRef]
  69. Wang, X.; Zhang, J.; Babovic, V.; Gin, K.Y.H. A Comprehensive Integrated Catchment-Scale Monitoring and Modelling Approach for Facilitating Management of Water Quality. Environ. Model. Softw. 2019, 120, 104489. [Google Scholar] [CrossRef]
  70. Pham, H.V.; Dal Barco, M.K.; Cadau, M.; Harris, R.; Furlan, E.; Torresan, S.; Rubinetti, S.; Zanchettin, D.; Rubino, A.; Kuznetsov, I.; et al. Multi-Model Chain for Climate Change Scenario Analysis to Support Coastal Erosion and Water Quality Risk Management for the Metropolitan City of Venice. Sci. Total Environ. 2023, 904, 166310. [Google Scholar] [CrossRef]
  71. Bello, S.F.; Wada, I.U.; Ige, O.B.; Chianumba, E.C.; Adebayo, S.A. AI-Driven Predictive Maintenance and Optimization of Renewable Energy Systems for Enhanced Operational Efficiency and Longevity. Int. J. Sci. Res. Arch. 2024, 13, 2823–2837. [Google Scholar] [CrossRef]
  72. Krishnan, S.R.; Nallakaruppan, M.K.; Chengoden, R.; Koppu, S.; Iyapparaja, M.; Sadhasivam, J.; Sethuraman, S. Smart Water Resource Management Using Artificial Intelligence—A Review. Sustainability 2022, 14, 13384. [Google Scholar] [CrossRef]
  73. Nickum, J.E.; Kuisma, S.; Bjornlund, H.; Stephan, R.M. Smart Water Management: The Way to (Artificially) Intelligent Water Management, or Just Another Pretty Name? Water Int. 2020, 45, 515–519. [Google Scholar] [CrossRef]
  74. Geissen, V.; Mol, H.; Klumpp, E.; Umlauf, G.; Nadal, M.; van der Ploeg, M.; van de Zee, S.E.A.T.M.; Ritsema, C.J. Emerging Pollutants in the Environment: A Challenge for Water Resource Management. Int. Soil Water Conserv. Res. 2015, 3, 57–65. [Google Scholar] [CrossRef]
  75. Quinlivan, L.; Chapman, D.V.; Sullivan, T. Applying Citizen Science to Monitor for the Sustainable Development Goal Indicator 6.3.2: A Review. Environ. Monit Assess 2020, 192, 218. [Google Scholar] [CrossRef] [PubMed]
  76. Hegarty, S.; Hayes, A.; Regan, F.; Bishop, I.; Clinton, R. Using Citizen Science to Understand River Water Quality While Filling Data Gaps to Meet United Nations Sustainable Development Goal 6 Objectives. Sci. Total Environ. 2021, 783, 146953. [Google Scholar] [CrossRef] [PubMed]
  77. Zhang, J.; Zou, T.; Lai, Y. Novel Method for Industrial Sewage Outfall Detection: Water Pollution Monitoring Based on Web Crawler and Remote Sensing Interpretation Techniques. J. Clean. Prod. 2021, 312, 127640. [Google Scholar] [CrossRef]
Figure 1. Integration of AI and PBM for Enhanced Water Quality Prediction.
Figure 1. Integration of AI and PBM for Enhanced Water Quality Prediction.
Water 17 01641 g001
Figure 2. Conceptual framework of AI-driven water quality management.
Figure 2. Conceptual framework of AI-driven water quality management.
Water 17 01641 g002
Figure 3. Overcoming Challenges: Prospects for Advancing AI in Water Quality Monitoring and Management.
Figure 3. Overcoming Challenges: Prospects for Advancing AI in Water Quality Monitoring and Management.
Water 17 01641 g003
Table 1. Comparison of Performance Indicators between AI-Based and Traditional Water Quality Monitoring Methods.
Table 1. Comparison of Performance Indicators between AI-Based and Traditional Water Quality Monitoring Methods.
Performance IndicatorsAI MethodsTraditional MethodsRef.
Timeliness
  • Enables real-time data acquisition and processing
  • Results delayed due to manual sampling and lab analysis
[10,11]
Accuracy
  • High precision through automated calibration
  • Susceptible to human error and environmental interference
[12,13]
Predictive Capability
  • Accurately predicts trends using DL models
  • Relies on simple models with limited forecasting capability
[11,14,15]
Anomaly Detection Capability
  • Detects sudden changes in real time via pattern recognition algorithms
  • Infrequent sampling delays anomaly identification
[16,17]
Decision Support
  • Provides real-time, data-driven recommendations
  • Depends on periodic reports and expert judgment
[18,19,20]
Data Integration
  • Automatically integrates data from multiple sources
  • Manual integration leads to fragmented data
[14,21]
Cost
  • Reduces long-term operational costs via automation
  • High recurring costs in labor, analysis, and maintenance
[22,23]
Table 2. Applications of AI Methods in Water Quality Monitoring. (Abbreviations: DO = Dissolved Oxygen; TDS = Total Dissolved Solids; ORP = Oxidation-Reduction Potential; EC = Electrical Conductivity; COD = Chemical Oxygen Demand; BOD5 = 5-day Biochemical Oxygen Demand; TN = Total Nitrogen; TP = Total Phosphorus; NH4+ = Ammonium nitrogen; TSS = Total Suspended Solids; WQI = Water Quality Index.).
Table 2. Applications of AI Methods in Water Quality Monitoring. (Abbreviations: DO = Dissolved Oxygen; TDS = Total Dissolved Solids; ORP = Oxidation-Reduction Potential; EC = Electrical Conductivity; COD = Chemical Oxygen Demand; BOD5 = 5-day Biochemical Oxygen Demand; TN = Total Nitrogen; TP = Total Phosphorus; NH4+ = Ammonium nitrogen; TSS = Total Suspended Solids; WQI = Water Quality Index.).
AI MethodsTaskData SourceParametersResultsRef.
SVM, ANNClassification; Source predictionIoT-based sensor collection systemTemperature, pH, Turbidity, DO, TDS, ORP, ECReal-time monitoring; Remote transmission; Self-feedback and self-regulation; Low cost; Portable[31]
K-MeansClassification; Automatic clustering analysis of sensor dataUnderwater robot (sensors)pH, Temperature, Turbidity, ECReal-time underwater monitoring; Remote control & data upload; Suitable for large water bodies; Sustainable deployment; Low cost, minimal maintenance, compact design[32]
RF, SVM, NNClassificationMultispectral sensorsEscherichia coli (E. coli)High accuracy[16]
PSO-SVRWQI predictionHyperspectral RS, In-situ dataComprehensive water quality parameters (pH, DO, COD, BOD5, TN, TP, TDS, etc.)High-precision; Suitable for large-scale dry area water quality estimation[33]
LSTMReal-time monitoring (aquaculture)Buoy sensorsTemperature, Salinity, DO, Flow velocityLow cost; Real-time data; Short-term prediction[22]
CNNReal-time monitoringSmartphone camera-captured sensor imagesE. coliHigh accuracy (99.99%); Objective; Simple operation[34]
CNNWater image classification (pollution)IoT-collected water imagesPollution type visual features (e.g., texture, oil stains, animal carcasses, etc.)High precision; Enhanced features; Real-time monitoring/feedback[35]
Boosting-IPW-PLSWastewater monitoringSensorsCOD, TSS, Oil and GreaseHigh precision; Handles noise; Online monitoring[36]
ANNSensor fault detection and replacement controlSensorsNH4, pH, ORP, DO, TSSHigh accuracy; Rapid fault ID; Automatic control replacement after fault detection[37]
ANNPrediction and disinfectant optimization controlLaboratory monitoring + sensorsE. coliHigh-precision; Optimizes peracetic acid disinfectant usage[38]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Zou, S.; Ju, H.; Zhang, J. Water Quality Management in the Age of AI: Applications, Challenges, and Prospects. Water 2025, 17, 1641. https://doi.org/10.3390/w17111641

AMA Style

Zou S, Ju H, Zhang J. Water Quality Management in the Age of AI: Applications, Challenges, and Prospects. Water. 2025; 17(11):1641. https://doi.org/10.3390/w17111641

Chicago/Turabian Style

Zou, Shubin, Hanyu Ju, and Jingjie Zhang. 2025. "Water Quality Management in the Age of AI: Applications, Challenges, and Prospects" Water 17, no. 11: 1641. https://doi.org/10.3390/w17111641

APA Style

Zou, S., Ju, H., & Zhang, J. (2025). Water Quality Management in the Age of AI: Applications, Challenges, and Prospects. Water, 17(11), 1641. https://doi.org/10.3390/w17111641

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop