Advances in Remote Sensing and Sensor Technologies for Water-Quality Monitoring: A Review

Abstract

Water-quality monitoring plays a vital role in protecting and managing water resources, maintaining ecological balance and safeguarding human health. At present, the traditional monitoring technology is associated with risks of low sampling efficiency, long response time, high economic cost and secondary pollution of water samples, and cannot guarantee the accuracy and real-time determination of monitoring data. Remote sensing (RS) technology and sensors are used to automatically realize the real-time monitoring of water quality. In this paper, the principles and composition of remote monitoring systems are systematically summarized. For the RS technology, indicators including chlorophyll-a, turbidity and total suspended matter/solids, colored dissolved organic matter, electrical conductivity (EC), dissolved oxygen (DO), temperature and pH value were considered, and for sensors monitoring, the parameters of pH value, temperature, oxidation reduction potential, DO, turbidity, EC and salinity, and total dissolved solids were analyzed. The practical applications of remote monitoring in surface water, marine water and wastewater are introduced in this context. In addition, the advantages and disadvantages of remote monitoring systems are evaluated, which provides some basis for the selection of remote monitoring systems in the future.

1. Introduction

Due to increasing levels of domestic use and industrial activities, a large amount of domestic sewage and industrial wastewater has been discharged into the water environment, resulting in frequent water-pollution incidents. The quality and availability of freshwater resources have become a key challenge facing the world [1]. According to the statistics, the global per capita availability of fresh water dropped from 17,000 m³ in 1950 to 6300 m³ in 2015, and is expected to drop to 4800 m³ by 2025. Therefore, there is an urgent need to monitor the quality of water in a series of contexts, such as rivers, lakes, groundwater, oceans and wastewater [2,3,4].

Water-quality monitoring plays a vital role in protecting and managing water resources, maintaining ecological balance and safeguarding human health. Commonly used water-quality parameters can be divided into three categories, including physical, chemical and biological parameters [5]. Among them, temperature, color, turbidity and odor are the most basic physical parameters. Chemicals are key factors affecting water quality, so pH, dissolved oxygen (DO), organic carbon, suspended solids and total solids should be continuously monitored in real time. In addition, some biological indicators can predict the quality of water. It is worth noting that physical, biological and chemical parameters also interact with each. For example, the amount of suspended solids can affect the turbidity of water, while temperature and chlorophyll affect the amount of flora [6].

At present, the traditional monitoring technology requires manual sample collection, preservation and preparation, and includes wet chemical treatment to achieve experimental testing and analysis, ultimately with the help of precision instruments [7]. It is associated with the risks of low sampling efficiency, long response time, high economic cost and secondary pollution of water samples, and cannot guarantee the accuracy and real-time determination of monitoring data [1], which makes it difficult to meet the comprehensive and dynamic needs of water-quality monitoring. In order to overcome these shortcomings, the remote-data water-quality monitoring system came into being. Remote sensing (RS) technology and sensors are used to automatically realize real-time monitoring of water quality in oceans, surface water and accident wastewater because of the agility of the platform, including real-time acquisition and sufficient accuracy [8,9].

Data acquisition, as the core of the remote-data water-quality monitoring system, has become a concern in the research. RS and sensor are the two most studied acquisition methods; they mainly differ as to the presence or absence of direct contact with the water. RS technology has high spatial and temporal resolution and a high coverage rate, and can realize rapid, synchronous and continuous monitoring of large-scale, multi-level and multi-dimensional water bodies. It can reflect the overall and local quality status of water bodies, and is suitable for macro and dynamic monitoring. Moreover, a variety of spaceborne sensors at visible, infrared and microwave wavelengths have been used to monitor water quality [10]. In the microwave field, radiometers and synthetic aperture radars are effective and feasible solutions for estimating the temperature and salinity of the sea surface. Meanwhile, a series of spectral RS data have been proved successful and widely used in water-quality inversion in the visible and infrared fields [11]. Monitoring data from Landsat-9, launched in September 2021, are now available to users around the world. However, RS technology is susceptible to the interference of external environmental factors, so complex atmospheric correction and water color correction are needed. Establishing a suitable inversion model and algorithm is the effective way to achieve high-precision monitoring. For example, Cai et al. [12] utilized HY-1C, HY-1D and GF-1 Wide Field View (GF-1 WFV) satellite data to establish the Qiantang River Tidal Surge Index (QRI), aiming to precisely determine the location and details of the Qiantang River tidal surges. The results indicated that QRI proved to be an effective extraction method and had potential applicability in similar tidal surge lines across different regions. Bu et al. [13] built a new chlorophyll-a (Chl-a) reversal model (PGS-C) based on GF-4 geosynchronous optical satellite data to analyze the distribution of Chl-a in the Yangtze River estuary, and the results showed that the correlation coefficient was 0.9123. The modeling value was highly consistent with the field value, indicating that the model had strong adaptability.

Sensor technology can realize the direct measurement of a variety of water-quality parameters with high sensitivity and accuracy, and can reflect the vertical and horizontal mass distribution of water. For instance, Jayaraman et al. [14] employed a genetic algorithm to optimize the hybrid model of support vector machine (GA-SVM), using 5000 pieces of data to predict the trend of a water-quality index from 2018 to 2023. This model demonstrated extremely high prediction accuracy (RMSE = 0.04474, R² = 0.96580) and reliability, providing an effective technical tool for water-quality management. Matos et al. [15] designed a cost-effective optical sensor using infrared backscattering, nephelometry and optical transport technology. It has been proven to monitor the turbidity and concentration of suspended solids in ocean water for 22 d continuously, and further proven to have a wide dynamic range and high accuracy. However, the sensor needs to be calibrated and maintained regularly. The existence of biological fouling in the ocean will cause some deviation, and manual regular maintenance is still a major factor limiting the wide application of the technology.

In recent years, numerous comprehensive review articles have explored the application of RS technology in water-quality monitoring [10,16,17,18,19,20]. They have focused on data sources, inversion algorithms, or specific water bodies. Compared with these studies, the innovation of this review lies in the following: (1) integrating RS and sensor technologies into a unified framework for remote water-quality monitoring and systematically comparing and discussing the principles, monitoring parameters, application examples, and advantages and disadvantages of both technologies, thus providing researchers with a clear foundation for selecting the appropriate technology; (2) offering an engineering application perspective that not only examines inversion algorithms but also completely details the architecture, hardware carriers and communication protocols of the monitoring system, an approach that holds significant practical value; and (3) expanding the application scenarios, in that, in addition to natural water bodies, the analysis also particularly focuses on the application status and challenges of sensing technology in industrial wastewater treatment, demonstrating the potential of the technology in pollution control.

2. Remote Data Acquisition and Treatment Systems

Remote data acquisition in the context of treatment systems has been widely used for its adaptability and high levels of accuracy. For the sake of brevity, the reviewed remote water-quality monitoring systems are summarized and compared in

Table 1. Comparisons of different remote data acquisition and treatment systems.

Data Acquisition	Carrying Platform	Wireless Interface	Controller	Data Storage	Water Body	Battery Energy	References
RS	UVA				Yangtze River		[21]
	UAS				Tharsis mine site		[22]
Sensors	USV	LoRa	Arduino	TTN	Gambang Lake	Solar	[23]
		LoRa	Esp32 microcontroller	Antares	Citarum River		[24]
		LoRa	Arduino	TTN	Canal de Burguera		[25]
		LoRa	Arduino	MySQL	Arno River		[26]
		LoRaWAN	Packetduino	Network Server	Fish pond		[27]
		GPRS /GSM	Waspmote V1.2	FLASH memory	Fiji surface water	Solar	[28]
	Amphibious UAV	4G	Raspberry Pi/Arduino pro	Google Firebase	A lake near Ambattur		[29]
	Unmanned Boat	WiFi	Arduino	OneNET	Small-scale water area		[30]
	UVA	WiFi	Raspberry Pi/Arduino	SD card	Simulated experimental water		[31]
		HC-08 Bluetooth	Arduino MEGA 2560	264K bytes RAM	Chia-Ming Lake		[32]
		GSM/ZigBee	LPC 2148	Central server	Simulated experimental water		[33]

.

2.1. Composition of the System

The remote data acquisition and treatment system consists of a data acquisition layer, data transmission layer, data storage layer and data processing layer. For instance, Jabbar et al. [23] designed a new Long Range Wide Area Network (LoRaWAN)-based IOT system for water-quality monitoring in rural areas (Figure 1a). Firstly, four sensors measured the water-quality parameters and transferred the information to the multi-sensor smart node, which consisted of an Arduino Uno microcontroller-based board and LoRa shield. The microcontroller was used to gather data from different sensors and periodically sent the data as a payload stream over the 915 MHz Cytron LoRa radio to the LoRaWAN gateway. Then, the information gathered was transmitted by the LoRaWAN gateway to the cloud, specifically, The Things Network (TTN). TTN was integrated with the ThingSpeak IoT platform and ThingView mobile apps, which could store data. The water-quality information was available on both web-based and mobile-based dashboards, and the users could easily access the water-quality parameters by using their internet-connected devices. Mamun et al. [28] also presented a Smart Water Quality Monitoring System (SWQMS), which has been developed and successfully deployed in five Fijian locations relative to the aforementioned key performance indicators (KPIs) (Figure 1b,c). The system comprised a Waspmote V1.2 microcontroller board, water sensors, Subscriber Identity Module (SIM)-900, General Packet Radio Service (GPRS)/Global System for Mobile Communications (GSM) module and a Geographic Information System (GIS) for interface and monitoring. The SWQMS also has an Analog to Digital Converter (ADC), a data storage card. Similarly, the two systems all have the four parts described at the beginning of this paragraph.

The data acquisition layer is mainly composed of RS and sensor equipment, and can realize the automatic, real-time and continuous collection of water-quality parameters [34]. RS includes satellite RS, aerial RS, unmanned aerial RS, and so on. Optical, radar, infrared and other sensors can be used for RS inversion relative to color, temperature, plankton, suspended matters, Chl-a and other parameters [35]. Sensors can be installed in the forms of fixed water-quality monitoring stations, mobile-monitoring vehicles or ships, and underwater drones, and can be used to monitor the pH value, electrical conductivity (EC), dissolved oxygen (DO), ammonia nitrogen (NH₄⁺-N), total nitrogen (TN) and total phosphorus (TP) of the water online. Sensors should be selected according to their characteristics, the monitoring objects, and the desired frequency and accuracy of the water measurements. In addition, the positions of the sensors should be as far apart as possible to avoid the interference of water flow. The data transmission layer is mainly composed of the IoT technology, which is responsible for transmitting the collected data to the data storage layer via wired or wireless transmission. Data transmission equipment should accomplish the functions of data encryption, compression and verification to ensure data security and integrity. Relevant wireless communication technologies include GPRS, 3G (3rd Generation Mobile Communication Technology), 4G, 5G, WiFi, Bluetooth, ZigBee, LoRa, NB-IoT, and so on. Usually, based on the data volume, transmission distance, power consumption and other requirements, researchers can choose an appropriate communication protocol and network. The data storage layer is mainly composed of cloud computing technology, which is responsible for storing the data from the data transmission layer in the cloud server or local database to achieve safe, reliable and efficient data storage. After considering the security, privacy, cost and other needs associated with the different types of data, choosing the right cloud service provider and cloud service model is straightforward, within reason. The data processing layer is responsible for cleaning, integrating, analyzing, mining and displaying the data in order to realize the value, intelligence and visualizations relating to the data. Statistics, machine learning, deep learning and other methods can be used to describe, predict, diagnose and recommend aspects of the water-quality data [36]. Similarly, neural networks, convolutional neural networks, recurrent neural networks and other models can also be used for semantic understanding, image recognition, and voice interactions relevant to the water-quality data. In addition, there are also systems equipped with remote monitoring platforms. These systems have functions including data alarms, notifications and push, which can bring attention to the occurrence of abnormal water quality in sufficient time.

2.2. Data Acquisition

Data acquisition, as the basis of a remote data acquisition and treatment system, determines the scope, precision and real-time elements of water-quality monitoring. There are two ways of data acquisition: RS and sensors.

2.2.1. RS

RS technology refers to the obtaining of data such as spectra and images of water bodies by the use of RS instruments carried by aircrafts or satellites, without coming into contact with the water bodies. Subsequently, water-quality parameters can be obtained by deduction. In recent years, data monitoring sources have progressed, moving from satellites and airborne applications to UAV and water-surface buoy data [37]. The water-quality parameters monitored by RS technology have gradually transitioned from monitoring Chl-a to calculating more than ten chemical parameters. The application range has been progressively developed from optical parameters to non-optical parameters. Non-optical sensors such as radar and infrared sensors can pick up electromagnetic signals in the microwave range emitted from the surface of the earth. Between 0.7 and 3.0 wavebands, water has poor reflectance in RS. This portion of the electromagnetic spectrum matches most closely to Band 7 on the Landsat MSS sensor, Band 4 on the Landsat TM sensor (0.76–0.9), Band 3 on the SPOT-HRV sensor (0.79–0.89) and Band 2 (0.72–1.1 m) on the (National Oceanic and Atmospheric Administration) NOAA A Very High Resolution Radiometer (AVHRR) series. All of these sensors have been shown to be extremely useful for imaging open water zones at these wavelengths. The research relating to the monitoring algorithms has made rapid progress, from regression and correlation algorithms to artificial intelligence and machine learning algorithms [38]. Therefore, the efficiency and accuracy of RS are improving constantly.

2.2.2. Sensors

Sensor monitoring technology refers to instruments in contact with the water that obtain water signals and other data through use of of a variety of sensors, which are based on chemical, physical, biological and other principles, so as to convert the data into water-quality parameters. According to the research, water-quality monitoring technology based on fixed monitoring systems has been widely studied. For example, Pramana et al. [39] designed a water-quality monitoring system with an early warning system that can be monitored remotely in real time using an internet-enabled smartphone or computer application. However, the limitation of a fixed monitoring system is that it can only monitor fixed locations [40].

Remote water-quality monitoring methods have seen advancements with the integration of robots equipped with sensors and samplers in environmental RS research [41]. The emergence of robots solves the problem of manually locating nodes. At the same time, these devices can monitor a large range of low altitude areas quickly, bridging the gap between field monitoring and traditional space-based RS [42]. Drones, unmanned vehicles, submersibles and boats are already being used for remote monitoring of water quality. Esakki et al. [29] combined the characteristics of multi-rotor UAV with a hovercraft to design an amphibious vehicle. A prototype was built with a 7 kg payload capacity and successfully tested for stable operations in flight-based and water-borne modes. IoT-based water-quality measurement is performed in a typical lake and water quality is measured using pH, DO, turbidity and EC sensors. The developed vehicle was expected to meet the functional requirements of disaster missions, focusing on the water-quality monitoring of large water bodies. However, most of the existing systems focus on a single function selected from among obstacle avoidance, water-quality monitoring and surface cleaning. Chang et al. [32] developed a multi-functional unmanned ground vehicle (MF-USV) based on sensor fusion technology; the device can simultaneously achieve the functions of obstacle avoidance, water-quality monitoring and surface cleaning. In addition, these research results are also closely related to the data transmission mode of the entire system. The possible communication distances for WiFi and ZigBee are short, and the anti-interference ability is weak. Also, the power consumption of 3G, 4G, 5G and GPRS applications is very high, and additional charges may be incurred. It is difficult to achieve continuous real-time water-quality monitoring effectively. LoRa stands out for its low cost, long communication distance and high endurance. All in all, sensors play an essential part in water-quality monitoring. Compared with fixed monitoring systems, mobile monitoring systems that are based on robots have prospects that are more broad.

2.3. Data Storage

Data storage is the key part of a remote data acquisition and treatment system, one which determines the safety and reliability, as well as the high efficiency, of a water-quality data system. Data storage mainly includes cloud computing technology and local database technology. According to the research, cloud computing technology has been widely used in storing data.

Cloud computing technology is a technology that uses the internet to store data on cloud servers distributed in different locations to achieve dynamic distribution, elastic expansion and on-demand use of data. Cloud computing technology can realize safe backup, fast access and low-cost maintenance of water-quality data, and is suitable for large-scale, distributed and multi-source storage of water-quality data. However, due to the impacts of internet bandwidth, stability, security and other factors, there may be delays in the transmission of water-quality data, loss, leakage and other risks. This approach can be combined with encryption, authentication, authorization and other technologies to assure sufficient security. At present, many researchers prefer to use cloud computing services provided by the cloud service providers (CSPS) available to the public; with these services, users can access and use the resources of cloud servers through the internet, and pay according to their usage.

Surprisingly, we found that there was little in the literature about RS that introduced the question of data storage specifically. Due to the fact that the accuracy of an RS system is correlated with data treatment, the models and algorithm are the sections thought most valuable. However, data storage is usually involved, whether detailed or simple. Take, for example, a water-quality monitoring and filtration system designed for Malaysia, in which the data processed by the microcontroller unit was uploaded and updated into Cloud storage and the ThingSpeak application through a data communication unit, the ESP8266 Wi-Fi module [43]. After logging into the ThingSpeak application with our unique user ID and password, we could access data from the monitoring system in real time, provided that the system was connected to the internet.

2.4. Data Treatment

Data processing refers to the obtaining of the required water-quality information through cloud computing, big data, artificial intelligence and other technologies, applications which can clean, integrate, analyze and mine the data. For RS and sensor-based technology, the stages of data treatment are basically the same. In both approaches, pretreatment, feature extraction and data presentation are included.

First of all, it is necessary to accomplish data preprocessing. Atmospheric correction, geometric correction and radiation correction are performed on the RS images, which can eliminate the interference of dark noise and stray light, and order the diffracted light within the images. These are regarded as efficient ways to improve the quality and accuracy of the data.

Subsequently, in a more complex process, the RS data is used to extract parameters and establish models. Specifically, the optical parameters of water, such as chlorophyll, suspended matter and colored soluble organic matters, will be extracted [44]. Meanwhile, based on the spectral, spatial and temporal features of the RS images, mathematical models appropriate to the water-quality parameters and spectral reflectance are established to carry out the inverse calculations necessary for the water-quality parameters. With reference to the array of water-quality parameters obtained by RS-based monitoring, many mathematical models interfacing between the RS data and the spectral reflectance of the water have been established. In recent years, machine learning for remotely sensed water-quality estimation has become popular thanks to the advances in algorithm development, computing power, sensor systems and data availability. Currently, the main methods for building water-quality parameter prediction models include the partial least square regression algorithm (PLS), artificial neural network (BP), convolutional neural network (CNN), random forest algorithm (RF) and machine learning [45]. Deep learning architectures, including recurrent and convolutional neural networks, appear particularly attractive, due to their great success in many recent studies, outperforming many other methods in a variety of RS applications [46]. Taking the water quality of Hong Kong’s coastal waters as an example, a study evaluated the application potential of machine learning technology in RS data processing. Determinations of SS, Chl-a and turbidity were estimated with several machine learning techniques, including artificial neural network (ANN), random forest (RF), cubist regression (CB), and support vector regression (SVR). Among them, the highest accuracy as to the water quality was achieved by ANN. Based on images from an M600 Pro UAV manufactured by DJI Lnc, Wei et al. [47] took a polluted river as an object of study. Surprisingly, compared with the semi-empirical model, the accuracy of the PSO-LSSVM algorithm in predicting the SSC was significantly improved.

Finally, the retrieved water-quality parameters will be processed by spatial interpolation, classification, statistical analysis, etc. Then, water-quality distribution maps, water-quality evaluation maps and water-quality change maps are generated to display the spatial distribution and dynamic changes in water quality in an intuitive way.

3. RS Technology

3.1. Fundamentals of RS Technology

The application of RS technology in water-quality monitoring is based on its ability to obtain long-distance and large-scale surface information. The monitoring and assessment of water quality and environmental conditions can be realized by RS technology, which has been proved to be an important tool for the study of spatial and temporal information associated with water quality [48,49]. Its basic principle involves the interactions between electromagnetic waves and water bodies, RS data acquisition, data processing and analysis and water-body parameter inversion. Specifically, RS technology can utilize the characteristics of electromagnetic waves of different wavelengths interacting with water bodies to infer water quality. When electromagnetic waves pass through or come into contact with water bodies, phenomena such as reflection, transmission and absorption will occur, and water bodies of different compositions have different effects on electromagnetic waves. Dissolved organic matter, suspended matter, algal bloom and other substances will affect the propagation and reflection of electromagnetic waves in water, thus reflecting the optical properties of the water body and its degree of pollution [50].

Additionally, RS technology can acquire data through sensors carried by satellites, airplanes and other platforms. These sensors cover visible, infrared and thermal infrared bands, which can capture the spectral information and temperature distribution on the surface of the water body [51]. Through the acquisition of data in different bands, it is possible to realize the monitoring of water bodies as to many of their aspects, such as water color, transparency, chlorophyll content, etc. After the effective RS data is obtained, it needs to be processed and analyzed in order to transform it into useful information. With professional data processing software, the water reflectivity, turbidity, chlorophyll content and other indicators can be extracted, and then these indicators can be analyzed to assess the status of the water quality, so as to find out the sources of pollution and the change trends, in order to provide a scientific basis for the management of the water environment.

Finally, with the help of reflectivity modeling and a water-quality RS inversion algorithm, the RS data can be transformed into water-body parameters such as chlorophyll concentration, transparency, dissolved organic matter content, etc., so as to realize the quantitative assessment of the water-quality condition [52]. This parameter inversion method provides a quantitative means for water-quality monitoring, which can more accurately reflect the degree of pollution and ecological changes in water bodies. For example, Figure 2 demonstrates the model structure used for estimating water-quality parameters in wetlands. In general, the basic idea of using RS technology for water-quality evaluation is to establish the inverse model of water-quality indicators and then derive the spatial distribution of the water-quality indicators, based on the mode [18]. By applying the above ideas, rapid and accurate monitoring of water quality and environmental conditions can be realized, providing important support for water resource management and environmental protection.

3.2. Water-Quality Indicators of RS Technology

RS technology can quantitatively estimate water-quality parameters by analyzing the significant influence of water-quality components on spectral characteristics. These parameters include optically active constituents (OACs) and non-optically active constituents (NOACs). OACs are the category most effectively monitored by RS technology, as they directly alter the optical properties of water bodies, thereby influencing the reflection spectra at specific wavelengths. Chl-a, suspended particulate matter (SPM) and colored dissolved organic matter (CDOM) are the most crucial OACs in water color RS [53], interacting with light through absorption, refraction and scattering. In contrast, NOACs such as DO, chemical oxygen demand, biochemical oxygen demand and total nitrogen lack significant optical properties and cannot be directly measured using spectral methods [54]. However, changes in NOACs are often closely related to OACs or the thermodynamic processes of the water bodies, and can be estimated by establishing indirect models.

3.2.1. Chl-a

Chl-a is an important indicator used in evaluating the degree of eutrophication in water bodies [55]. Its concentration directly influences the optical properties of the water body, thereby providing an important basis for the use of RS technology. RS technology can realize the monitoring and evaluation of the degree of eutrophication of water bodies by monitoring the Chl-a absorption characteristics in the reflectance spectra on the surface of the water bodies and extrapolating the chlorophyll-a concentration in the water bodies [56,57]. In addition, Chl-a is also the most extensively studied parameter in RS technology [58]. It exhibits absorption peaks in the range of 430–700 nm, as well as near 700–720 nm, due to fluorescence effects and cell scattering, which forms the physical basis for most inversion algorithms [54].

Currently, the primary inversion methods include empirical models, semi-analytical models and machine learning algorithms. Widely used empirical algorithms include the dual-band ratio model (such as Rrs(708)/Rrs(665)) [59] and the normalized difference chlorophyll index (NDCI) [60]. These methods achieve rapid estimation by establishing a statistical relationship between the combined band reflectance and the measured concentration. They show strong applicability in multi-spectral satellite data, such as data from Sentinel-2, and can achieve an R² value of up to 0.845. Semi-analytical models commonly include three-band and four-band models [61]. However, despite the widespread use of empirical and semi-analytical methods, they still have limitations in terms of inversion accuracy, generalization ability, anti-interference capability and computational efficiency.

With the development of artificial intelligence, machine learning methods have been increasingly applied to the inversion of Chl-a, especially for nonlinear optimization problems. These methods effectively avoid the errors caused by atmospheric correction in traditional methods and demonstrate superior universality. Support vector machine (SVM) [62] and random forest (RF) [63] algorithms have shown particular promise in Chl-a inversion, as they can capture the nonlinear relationship between the spectrum and Chl-a in complex optical environments. For instance, a study that employed an SVM model to estimate Chl-a in multiple lakes in China achieved a prediction accuracy exceeding 0.88, significantly outperforming traditional empirical algorithms, thus demonstrating the applicability of these methods in complex water bodies [62].

By integrating multi-source RS data, these algorithms can effectively support the spatiotemporal dynamic monitoring of water-quality parameters in eutrophic water bodies.

3.2.2. Turbidity and Total Suspended Matter/Solids (TSM/TSS)

Turbidity is an indicator of the contents of suspended and particulate matter in a water body, a factor which affects the transparency and clarity of the water body, underwater light field distribution and nutrient transport [64]. RS technology can be used to monitor and assess the turbidity of water bodies by obtaining reflectance-based spectral information from the surface of the water body, analyzing the optical properties, and extrapolating the turbidity of the water body. Suspended matter includes inorganic sediments, clay, and organic debris, and its concentration typically increases the reflectance of water bodies in the red and near-infrared regions. Specifically, a reflection peak forms at 750 nm, while an absorption trough appears at 950 nm. As the concentration increases, the reflection peak shifts toward longer wavelengths, and the absorption trough moves toward shorter wavelengths in the blue region [65]. This optical response provides the theoretical foundation for developing RS inversion models.

Currently, the main inversion methods are classified into three categories: empirical models, semi-analytical models and machine learning algorithms. Empirical models, such as single-band index or band combination regression, rely on statistical relationships to achieve rapid estimation. For example, using the 645 nm band of MODIS-Aqua or the combination of the green, red and blue bands of Landsat to establish regional regression models is straightforward but has relatively low universality [66]. The semi-analytical methods are based on bio-optical models used to separate the contributions of particulate matter absorption and backscattering from the RS reflectance. For instance, the two-step inversion algorithm based on GOCI data has significantly improved accuracy and applicability in complex water bodies such as Taihu Lake and Hangzhou Bay [67]. In recent years, machine learning algorithms such as RF, XGBoost and extreme learning have gradually become mainstream [68]. These algorithms are capable of capturing complex nonlinear relationships between spectra and water-quality parameters, effectively overcoming the dependence of traditional models on atmospheric correction and empirical parameters. Studies have shown that machine learning models optimized by genetic algorithms, namely, particle swarm optimization (such as GA-RF, PSO-BPNN) can further improve inversion accuracy and generalization ability [69]. For example, XGBoost has shown better stability than traditional ratio models in Sentinel-2 image applications, and the optimized BPNN model has further reduced the TSM inversion error to 4.04 mg/L [70].

When combined with multi-source RS data, these methods have significantly advanced the development of TSM/TSS monitoring, transitioning from regional empirical models to large-scale RS quantitative inversion with physical mechanism support, high accuracy and strong adaptability. They have provided crucial technical support for water environment management, sediment transport research and drinking water safety.

3.2.3. Colored Dissolved Organic Matter (CDOM)

CDOM is an organic substance in water bodies, originating from biological decomposition, plant residues, human activities, and so on, a substance that has an impact on the color and transparency of water bodies [71]. As an important component of dissolved organic carbon (DOC) [72], CDOM can significantly absorb ultraviolet and blue light wavelengths, causing a decrease in water transparency and directly affecting the process of photosynthesis and the energy transmission in aquatic ecosystems [73].

RS technology enables the monitoring and assessment of organic matter in water by detecting specific spectral absorption features of DOM in surface water reflectance spectra, and then inferring the DOM concentrations [49]. Specifically, CDOM exhibits an exponential decrease in absorption coefficient across the 250–700 nm wavelength range as wavelength increases. Quantitative indicators often use absorption coefficients at 355 nm or 440 nm [74]. RS technology achieves the spatial inversion and dynamic monitoring of CDOM concentration by exploiting its absorption characteristics in specific spectral bands.

Currently, the RS inversion methods of CDOM can be classified into three categories. Empirical and semi-empirical methods estimate CDOM by establishing statistical relationships between multi-band reflectance combinations and concentration. For example, ratio-based regression models using green and red bands are common. While computationally simple, these approaches are susceptible to interference from suspended particles, chlorophyll and other water constituents [75]. Semi-analytical methods, which are based on bio-optical models and radiative transfer theory, decouple the absorption contribution of CDOM from RS reflectance using advanced algorithms such as the quasi-analytical algorithm (QAA). These methods possess clear physical mechanisms and maintain high accuracy, even in turbid waters. For instance, the QAA_cj model achieved a correlation coefficient of 0.90 and a root mean square error of 0.07 m⁻¹ while estimating the CDOM absorption coefficient at 443 nm in the Yangtze River estuary [76]. Machine learning methods, such as RF, XGBoost and support vector regression, can capture the complex nonlinear relationship between spectra and CDOM concentration [77]. They perform well in terms of inversion accuracy and model generalization ability. For example, the RF model has a determination coefficient (R²) of 0.85 for the absorption coefficient at 355 nm [78]. When combined with optimization methods such as genetic algorithms or principal component analysis, the robustness and cross-scenario applicability of the model can be significantly improved [79,80].

In summary, RS inversion-based methods relevant to CDOM are evolving from traditional empirical approaches toward methods integrating radiative transfer theory and data-driven modeling. The combination of multi-source RS data and intelligent algorithms is expected to improve inversion accuracy and enhance model adaptability relative to diverse water optical types, thereby supporting effective water-quality monitoring and management.

3.2.4. Non-Optically Active Constituents (NOACs)

Currently, research on NOACs remains relatively limited. Due to the weak spectral response and the subtle characteristics of NOACs, directly estimating their concentration from spectral data presents significant challenges. Therefore, most existing studies have turned to indirect estimation methods. First, the parameters of OACs, such as Chl-a, TSM and CDOM, are retrieved using the spectral reflectance data associated with water bodies. Then, based on the intrinsic correlation between OACs and NOACs, the NOACs content is indirectly calculated through empirical/semi-empirical models or semi-analytical models [81,82].

Further research has revealed that in specific water environments, NOACs may be associated with multiple OACs, meaning that multiple bands in the spectral curve may implicitly contain information related to NOACs. Artificial intelligence models exhibit unique advantages in uncovering such complex correlations. Even without explicitly establishing an exact correspondence between NOACs and OACs, these models can extract implicit information related to OACs from overlapping spectral features, thereby constructing more accurate models for estimating NOACs [83,84]. Chen et al. [85] used a typical rural river as the research object, combined high-resolution multispectral images from drones and ground-measured data, and employed nine machine learning models to estimate total nitrogen (TN) and total phosphorus (TP). The results showed that the optimal CatBoost regression model achieved an R² value exceeding 0.9, fully verifying the potential of the application of artificial intelligence models.

It is worth noting that, regardless of whether a traditional model or an artificial intelligence model is used, the estimation accuracy of the method is highly dependent on the stability of the relationship between NOACs and OACs. This relationship is susceptible to environmental factors such as hydrology and climate, making it difficult for the models to be directly generalized to other water-body types [86]. Despite this limitation, the quantitative estimation of NOACs still has significant reference value and application potential in improving the water-quality monitoring system and assessing water-pollution risks. For example, RS technology can realize the monitoring and assessment of DO in water bodies by obtaining reflectance-based spectral information on the surface of water bodies, and combining this with parameters such as water-body chromaticity and temperature to deduce the concentrations of DO in water bodies. It can also monitor the thermal infrared spectral information of the water surface, invert the temperature distribution of the water body and achieve remote sensing-based monitoring and assessment of water temperature [87]. Although RS technology cannot directly monitor the pH value of a water body, it can indirectly determine the pH range of a water body by obtaining the reflectance spectrum information on the surface of the water body and analyzing the chemical substances in the water body that are related to the pH value.

3.3. Practical Applications of RS Technology

RS technology in water-quality testing can realize the monitoring and assessment of a variety of water-quality indicators, providing important technical support for water-quality management and protection. By combining RS technology and field monitoring means, the quality status of water bodies can be understood in a more comprehensive and timely manner, and the sustainable utilization and protection of water resources can be promoted [88,89]. RS technology has been used in the monitoring of surface water and marine water quality (

Table 2. Applications of RS techniques to the monitoring of surface waters and marine waters.

Area	Sensor	Monitoring Index	Parameter Concentration	Equation/Algorithm	R2	RMSE	References
Laguna Lake	Sentinel-2	Chl-a, TSM	Chl-a: 10–30 mg/L; TSM: 25–170 mg/L	Normalized difference chlorophyll index	_	_	[90]
Lake Balaton	Landsat 4/5, L7 ETM+, L8/9 OLI Level 2, Collection 2, Tier 1	Chl-a	5–260 μg/L	RF	0.86	8.16 μg/L	[91]
Saginaw River	Landsat 8 OLI, Sentinel-2 MSI	CDOM	3.29–17.86 mg/L	$\mathrm{Landsat}\text{-}8: CDOM\left(440\right)=40.75e-2.463x$; $x=Rrs\left(OLI3\right)/Rrs\left(OLI4\right)$; $\mathrm{Sentinel}\text{-}2: CDOM\left(440\right)=28.966e-2.015x$; $x=Rrs\left(MSI3\right)/Rrs\left(MSI4\right)$	Landsat-8: 0.86; Sentinel-2: 0.78	Landsat-8: 1.13 mg/L; Sentinel-2: 1.41 mg/L	[92]
Urban rivers	Hyperspectral imagery	Transparency	_	XGBoost	0.97	_	[93]
Midwestern United States	Landsat-8, Sentinel-2	Blue-green algae (BGA), Chl-a, fDOM, DO, and turbidity	BGA: 0.1–9.3 μg/L; Chl-a: 0.6–74.4 mg/L; DO: 0.1–19.7 mg/L; fDOM: 0.3–156.2 QSU; SC: 247.9–654.8 μs/cm; Turbidity: 2.0–131.1 FNU	Deep learning model (pDNN)	BGA: 0.91; Chl-a: 0.88; DO: 0.89; fDOM: 0.93 SC: 0.87; Turbidity: 0.84	BGA: 0.863 μg/L; Chl-a: 7.561 mg/L; DO: 1.806 mg/L; fDOM: 14.496 QSU; SC: 448.463 μs/cm; Turbidity: 5.190 FNU	[94]
Sanalona Reservoir	Landsat-8, Box–Cox transformations	TOC, Chl-a, and TDS	TOC: 3.8–8.2 mg/L; Chl-a: 0.1–10.9; TDS: 131.5–227.5 mg/L	$Box-Cox \left(TOC\right) = 1 + \left[\right({TOC}^{-0.5625}-1) / (-0.5625) ({5.47202}^{-1.5625}\left)\right]$; $Box-Cox \left(TDS\right) = 1 + \left[\right({TDS}^1-1) / (1) ({149.722}^0\left)\right]$; $Box-Cox (Chl-a) = 1 + \left[\right({Chl-a}^{0.901463}-1) / (0.901463) ({1.93691}^{-0.0985367}\left)\right]$	TOC: 0.90; Chl-a: 0.96; TDS: 0.95	TOC: 2.10; Chl-a: 5.67; TDS: 27.91	[95]
Chebara Reservoir	Sentinel-2A MSI, Landsat-8/OLI	Turbidity	_	$\mathrm{Sentinel}\text{-}2: {Rrs(B2/B3)}^2+Rrs\left(\right(B2/B3)$ $\mathrm{Landsat}\text{-}8: Rrs\left(\right(B3/B2)$	Sentinel-2: 0.75; Landsat-8: 0.75	Sentinel-2: 0.5 NTU; Landsat-8: 0.5 NTU	[96]
White River	Landsat 7 ETM+, Landsat 8 TIRS	Temperature	_	Air2stream	0.97	1.58 °C	[97]
Tyrrhenian coasts in Italy	Sentinel-2 MSI	Chl-a, TSM	Chl-a: 0.1–7.37 μg/L; TSM: 1–20 mg/L	EmpReg algorithm	Chl-a: 0.85 μg/L; TSM: 0.5 mg/L	Chl-a: 0.33 μg/L; TSM: 1.95 mg/L	[98]

Notes: R², coefficient of determination; RMSE, Root Mean Square Error; Rrs, RS reflectance; OLI3 and OLI4 are the 3rd and 4th bands of Landsat-8 (green, 561 nanometers) and (red, 655 nanometers), respectively; MSI3 and MSI4 are the 3rd and 4th bands of Sentinel-2 (green, 560 nanometers) and (red, 664 nanometers), respectively; fDOM, Fluorescent dissolved organic matter; B2, Blue bands; B3, Green bands.

).

3.3.1. Surface Water-Quality Monitoring

Surface water is one of the important water resources for human life and production activities, and its quality is directly related to human health and the protection of the ecological environment. RS technology plays an important role in the detection of surface water quality and can realize the comprehensive monitoring and assessment of surface water quality. Currently, RS technology has been widely used in the contexts of water turbidity and transparency [93,99], Chl-a concentration, algal bloom, CDOM, water temperature and thermal pollution monitoring [100,101].

Chl-a concentration in surface water can be monitored by RS technology, thereby determining the growth of algae in the water and realizing the monitoring of the spatial distribution and change trends associated with chlorophyll concentration in water bodies [102]. Li et al. [91] conducted a study on Lake Balaton in Hungary, integrating multi-temporal Landsat images with on-site data. They applied the RF model and combined over-sampling and seasonal time constraints to achieve high-precision Chl-a inversion. The R² value was 0.86, with the model effectively overcoming the problems of scarce high-value samples and mismatched data time series. This approach provided a reliable framework for long-term RS-based monitoring of lakes with large fluctuations and sparse sampling. Algal blooms can affect water ecosystems and water quality, and timely monitoring can prevent the water quality problems caused by excessive algal growth. RS technology can be used to monitor algal blooms in water bodies, including cyanobacteria and planktonic algae [103,104]. In addition, the CDOM content in surface water can also be monitored using RS techniques. Knowing the CDOM content helps in assessing the degree of eutrophication and the ecological quality of water bodies. Chen et al. [92] utilized the combined data from Landsat-8 and Sentinel-2 satellites to inversely calculate CDOM concentration based on its optical properties. Subsequently, by leveraging the significant correlation between CDOM and DOC (R² = 0.86), they established an indirect estimation model for DOC, successfully achieving high-precision estimation and dynamic monitoring of DOC concentration in the Saginaw River estuary of Lake Huron. In a study of Laguna Lake in the Philippines, Caballero et al. [90] used Sentinel-2 satellite images and employed the C2RCC processor to simultaneously invert the concentrations of Chl-a and TSM. They also used the NDCI index to monitor the cyanobacterial bloom. The research found that the concentrations of TSM and Chl-a significantly increased after a typhoon, confirming the effectiveness and application potential of medium-resolution RS technology in monitoring the rapid responses of multiple parameters of water quality during extreme climate events.

Of course, the NOACs in surface water have also been monitored through the combination of RS and algorithms, with practical applications resulting. By combining UAV hyperspectral images with the XGBoost regression algorithm, high-precision quantitative inversion of the transparency of urban river water bodies can be achieved (R² > 0.97), effectively overcoming the limitations of traditional RS technology regarding insufficient spatial and spectral resolution in narrow water areas [93]. Meanwhile, by utilizing multi-spectral RS data from Sentinel-2, various band combinations and regression models can be constructed to effectively invert six key water-quality parameters in river water bodies: DO, permanganate index, ammonia nitrogen, TP, TN and turbidity [105]. RS technology can also monitor the surface and vertical temperature distributions of water bodies, including seasonal variations and temperature gradients. Understanding the temperature of water bodies is important for water ecosystem and water-quality management. Thermal pollution is also one of the key factors affecting ecosystem and water quality in water bodies, and RS technology can be used to monitor the thermal pollution of water bodies, including the increase in water-body temperature caused by industrial wastewater discharge, the urban heat island effect and greenhouse gas emissions.

In general, RS technology has the advantages of convenience, efficiency, wide coverage and high resolution in surface water-quality detection, which provides important technical assistance for surface water-quality monitoring. Through remote monitoring and data analysis, it can provide a more comprehensive and accurate understanding of the water-quality status, timely detection of water-quality anomalies and protection of water resources and the ecological environment. The integration of RS technology and algorithms represents the mainstream trend for future water-quality monitoring. Continuous innovation in deep learning models, such as pDNN, can enable synchronous and high-precision estimation of multiple inland water-quality parameters [94]. Future models will significantly outperform traditional machine learning methods; these new models include multiple linear regression and support vector machine regression, approaches that will enhance practical applications in water-quality time series reconstruction and abnormal event detection.

3.3.2. Marine Water-Quality Monitoring

The oceans are the most extensive bodies of water on Earth, and their water quality has a significant impact on marine ecosystems and human life and production activities. RS technology plays an important role in the monitoring of marine water quality and can realize the comprehensive monitoring and assessment of the quality of marine water bodies [106]. At present, RS technology has been effectively applied in the contexts of ocean surface temperature, ocean circulation, marine biodiversity, oil pollution and ocean surface acidification. Sea surface temperature is an important parameter in the marine environment, one which can reflect the distribution of ocean heat and seasonal changes and provide data support for climate change research [107]. Remote monitoring of sea surface temperature can be realized by using RS techniques [108]. For example, Safarkhani et al. [109] used RS to study the fluctuation of sea surface temperatures in the Persian Gulf. In order to better understand the laws of ocean motion and study the process of an ocean’s dynamics, it is extremely important to monitor the ocean’s circulation system. RS techniques allow for more timely and accurate monitoring of information, including data describing the ocean surface’s wind fields, currents and eddies. In addition, marine biodiversity is one of the important indicators for assessing the health of marine ecosystems and guiding marine conservation and resource management efforts. RS technology can be used to monitor biodiversity in the oceans, including phytoplankton, zooplankton and fish. Chl-a, as an indicator of phytoplankton biomass, is a crucial variable for assessing the health and status of the ocean. Tilstone et al. [110] systematically evaluated the performance of the Sentinel-3 OLCI, MODIS-Aqua and Suomi-NPP VIIRS satellite sensors, based on the Chl-a measurement data obtained from the transatlantic passes of the satellites. The results show that the POLYMER algorithm, when based on Sentinel-3 OLCI, can achieve more accurate Chl-a inversion and provide higher levels of effective coverage in the Atlantic circulation area, offering a key insight into selecting algorithms for use with satellite remote sensing to monitor the ecological environment of oligotrophic sea areas.

Oil pollution is currently one of the causes jeopardizing the health of the marine ecosystem. Due to the strong mobility of oil substances, their trajectory in the water is irregular, which makes them relatively difficult to manage. After the incidence of water-quality oil pollution, the pollution will first lead to a decrease in the transparency of the water body, and then a thicker layer of oil film will be formed, covering the water surface. At the same time, due to the influence of the surrounding water and air humidity, the oil film will gradually spread. Therefore, it is very important to monitor and identify the oil pollution in monitoring water quality [111,112]. RS technology can detect oil pollution on the surface of the ocean, including crude oil spills, ship discharges and pollution events caused by oil field development, and timely detection and monitoring of oil pollution can help to ensure a timely response and reduce the harm to the ecological environment. For example, Mannino et al. [113] collected measurements from the continental margin area in the northeastern United States to validate a satellite algorithm for ocean color in aid of the inversion of colored dissolved organic matter absorption coefficients and spectral slopes.

In addition, RS can be used to monitor the degree of acidification of the ocean surface, reflecting the ocean’s uptake of carbon dioxide and changes in chemistry. Understanding ocean acidification can help assess the vulnerability of marine ecosystems and assist in the development of relevant conservation policies. Storms in the ocean have a great impact on marine water quality and marine ecosystems, and can cause problems such as turbidity of seawater and an increase in the amount of suspended substances [114]. RS technology can provide important information support for marine environmental protection and safety by monitoring the storm characteristics on the surface of the sea and providing early warning of marine storms. In summary, the application of RS technology in marine water-quality monitoring covers many aspects and can provide important support for marine environmental protection, resource management and climate change research. Through remote monitoring and data analysis, a more comprehensive and accurate understanding of the dynamic changes in the marine system can be achieved, providing a scientific basis for the protection of the marine ecosystem and the sustainable use of marine resources.

Currently, there are still limitations in the spectral simulation of sea areas for different nutrient statuses. Factors such as the accuracy of atmospheric correction, the complexity of water-body optics, the influence of water depth and the quality of data sets remain the main challenges in improving the accuracy of RS inversion for nearshore water quality determinations. [98]. Although the Sentinel-2 data is applicable to coastal environment monitoring, further systematic work is still needed in aspects such as atmospheric correction verification, optical property analysis and construction of standardized data sets, in order to enhance the robustness and accuracy of the inversion of water-quality parameters for these two types of water bodies.

3.4. Benefits and Limitations of RS Technology

3.4.1. Benefits of RS Technology

RS technology, as a comprehensive, advanced modern monitoring technology, has the advantages of being portable and efficient and offering wide coverage, high resolution, adaptability and the ability to monitor in real time to solve the limitations of the traditional field measurements. It has therefore become the most effective means for obtaining environmental spatial and temporal information [115]. Specifically, RS technology saves a substantial amount of time, in addition to labor costs, because it can provide extensive coverage and monitoring of water bodies through satellites, airplanes and other carriers without field sampling and testing, and RS technology can also provide the timely updating and monitoring of data, which makes the monitoring work more efficient. In addition, RS technology can realize the monitoring of a wide range of water bodies including different types of waters such as lakes, rivers and oceans, and can provide a comprehensive understanding of the pollution situation and change trends associated with water bodies. Additionally, RS technology can provide high-resolution image data, which enables researchers to observe the subtle changes in water bodies more clearly, and is conducive to accurately analyzing the water-quality situation and monitoring water-pollution sources. Finally, RS technology can realize the real-time monitoring of water bodies, assisting in the timely detection of water-quality anomalies, which is conducive to the taking of timely countermeasures to protect water resources and water ecosystems. Overall, RS technology is a widely implemented, integrated, cost-effective and robust water-quality monitoring method [116].

3.4.2. Limitations of RS Technology

With the development of satellite RS technology in China, its advantages in the field of environmental monitoring will be further realized and the prospects for its application will be broader. However, RS-based monitoring technology at its present stage still has limitations, such as difficulties in data analysis, large uncertainties and the inability to replace field monitoring. In terms of data, due to the large volume and complexity of RS data, professional RS technicians are required in order to process and analyze the data, and there is a certain degree of difficulty in interpreting and applying the data, so more professional workers and technical support workers are needed, an additional investment that significantly restricts the wider popularization of RS technology in the application of water-quality monitoring [117]. In addition, the weather conditions during the monitoring process will also have a great impact on the resulting quality of the monitoring data. For example, weather phenomena such as cloud cover, rain and snow can affect the acquisition and accuracy of RS data, so the effects of RS-based monitoring become uncertain when weather conditions are unstable in some areas. In addition, although RS technology can provide a wide range of monitoring data, it cannot completely replace field monitoring; some water-quality parameters still need to be sampled and tested in the field to obtain accurate data, so the combination of RS technology and field monitoring is required in order to achieve a more comprehensive understanding of the quality of water bodies.

4. Sensors Technology

4.1. Fundamentals of Sensors

Sensors are important devices for monitoring the water environment [118]. A sensor is usually composed of sensitive components and conversion components, and can sense the measured signals and convert them into a usable output signal according to a certain pattern [119]. The sensitive components can directly sense or respond to the measured aspect, and the conversion element converts the measured signal sensed or responded to by the sensitive element into an electrical or digital signal suitable for transmission or measurement. The generated electrical signal is processed by the sensor conversion circuit to ensure accurate and reliable measurement results are obtained [120]. The processed signal is transmitted to other parts of the system, such as control systems, displays, or data recording devices.

Water-quality sensors can be roughly divided into physical sensors, chemical sensors and biological sensors. Among them, physical sensors can be separated into optical sensors and resonant sensors [121]. Chemical sensors, mainly electrochemical sensors, are a relatively mature type of chemical sensors that have been widely studied in terms of technological development. The principle of electrochemical sensors is to convert the chemical changes generated by the tested substance into electrical signals based on its electrochemical characteristics, thereby completing the function of detecting the composition and content of the tested substance [122]. In the biological detection sensor, the electrodes used for detection have been biologically treated by solidifying various microorganisms, antibodies, organelles and enzymes within the biocatalysts, and can react with various detected chemicals. Then, electrochemical devices are used to selectively measure the generated or consumed chemicals, and the relevant changes are converted into electrical signals [123].

4.2. Water-Quality Indicators of Sensors

Water-quality monitoring relies on accurate measurements of a range of key physical, chemical and biological indicators. These indicators, such as pH, temperature, DO, oxidation-reduction potential (ORP), turbidity, conductivity and salinity, and total dissolved solids (TDS), comprehensively reflect the health, suitability and ecological balance of the water body. Traditional manual sampling and laboratory analysis methods are inherently subject to delay and cannot meet the real-time data demands of modern aquaculture, environmental early warning systems and smart water management scenarios. However, with the rapid development of sensing technologies, the Internet of Things (IoT) and data analysis algorithms, continuous, online and automated monitoring of these indicators is now possible, utilizing various sensors. Furthermore, sensor data can be calibrated, compensated and predicted using advanced algorithms such as linear regression and machine learning models, significantly enhancing the accuracy, reliability and intelligence of monitoring systems.

4.2.1. Value of pH

The pH value is a critical indicator of the acidity or alkalinity of water bodies, directly influencing the survival of aquatic organisms, the rate of chemical reactions and the toxicity of pollutants. According to the different sensor principles, pH sensors can be divided into electrochemical pH sensors, fluorescent pH sensors, and enzyme pH sensors [124]. Electrochemical pH sensors, which include glass electrode sensors and ion-sensitive sensors, can measure the pH of a solution based on the potential changes in the electrodes [124]. Fluorescent pH sensors use fluorescent dyes as indicators. By observing changes in the fluorescence intensity or emission wavelength of the dyes, the pH of the solution can be inferred [125]. When the dye binds or dissociates with hydrogen ions, the luminescence-based characteristics change, and the pH can be measured by detecting these changes [126]. Enzyme-based pH sensors measure pH by measuring the activity changes of specific enzymes under different pH conditions. When the substrate undergoes a reaction under the action of enzymes, the generated signal can be correlated with the pH value [127].

In addition to traditional sensors, algorithms are frequently employed for calibration and compensation to enhance the accuracy of low-cost sensors. For example, a simple linear regression algorithm has been shown to effectively correct the output of low-cost sensors. The approach involves comparing the readings of low-cost sensors with the measurement values from high-precision standard instruments, such as the YSI Professional Pro, to establish a linear correction model. This process significantly improves the accuracy of low-cost sensors, from 76% to 97%, and reduces the relative error to a range of 0.27% to 4% [128].

4.2.2. Temperature

Technologies such as thermal sensors or thermocouples have been employed in temperature sensors. Thermocouple sensors use the thermoelectric effects between different metal conductors to measure temperature. When the connection between two different metals is subjected to temperature changes, an electromotive force is generated, and the temperature is determined by measuring the voltage change caused by the temperature difference. Wang et al. [129] designed a flexible temperature sensor by connecting graphene fibers and platinum, based on the principle of thermocouples, attaining a sensitivity of 29.9% μV/°C. The sensor exhibited good cycling stability, repeatability and stability. Thermistor components can measure temperature based on the principle that resistance or capacitance values vary with temperature. Single layer/multi-layer graphene can be directly used as a temperature-sensitive material for temperature sensors. The high linearity flexible graphene temperature sensor prepared by Zhang et al. [130] has a linearity of 0.999 in the range of 30 to 100 °C, and has the characteristics of fast response, short recovery time and good repeatability, reliability, stability and flexibility. In addition, temperature sensors can also use optical principles to measure temperature changes. The medium inside the optical fiber changes with temperature, causing changes in refractive index, resulting in changes in the frequency and phase of the reflected light. By analyzing these changes, fiber optic sensors can obtain temperature measurement results with high accuracy and electromagnetic compatibility [131,132]. In addition, piezoelectric temperature sensors also utilize the piezoelectric effect, the principle that certain crystals generate charges when subjected to force [133]. Different types of temperature sensors are suitable for different application scenarios, and the selection of the appropriate sensor usually depends on measurement requirements, environmental conditions and cost considerations.

It is important to note that despite the use of advanced technologies, such as hyperspectral imaging, in water-quality monitoring, real-time sensor systems still heavily rely on basic parameters like temperature, pH and DO [134]. The high-frequency collection of these parameters provides essential data support for water-quality assessment. Therefore, continuously optimizing the sensing performance of fundamental parameters such as temperature and promoting their deeper integration with information and communication technologies is of significant practical importance for developing an efficient and intelligent water-quality monitoring network.

4.2.3. Oxidation Reduction Potential (ORP)

The detection of ORP by sensors is based on measuring the potential difference between oxidants and reducing agents in the solution. ORP electrodes can reliably measure ORP in almost all aqueous solutions, and are usually not affected by color, turbidity, colloidal substances, or suspended solids in the solution. Lee et al. [135] prepared an ORP potential in situ sensing microelectrode array, one which can be used for in situ monitoring of ORP in water environments. ORP sensors are widely used in water-quality monitoring, wastewater treatment, bioremediation, industrial process control for evaluating environmental conditions, the monitoring of the removal processes of pollutants, and the control of chemical reactions [136]. High-quality ORP sensors should have fast responses and long-term stability in order to provide accurate measurement results in dynamic environments [137]. Moreover, the ORP sensor can detect various pollutants in both dissolved and particulate forms. When tested, it has identified the presence of all the pollutants under study, except for Escherichia coli [138].

A strong positive correlation exists between ORP and free chlorine concentration, making ORP a crucial proxy parameter for monitoring and evaluating disinfection efficacy and microbial safety in drinking water. ORP serves as an indicator of the oxidative capacity of water, reflecting its ability to purify or degrade waste compounds. Water with an ORP value above 300 mV is considered strongly oxidative [139]. Since free chlorine is a potent oxidizing agent, its concentration directly influences the ORP reading. Therefore, monitoring changes in ORP can rapidly, although indirectly, signal the diminishment of the free chlorine concentration [140].

However, ORP sensors have certain limitations in practical applications. In systems using chloramine as the disinfectant, the response of ORP to concentration changes is less pronounced and consistent, limiting its standalone applicability. Additionally, ORP signals may exhibit a delayed response after free chlorine is reintroduced, posing challenges for direct use in real-time closed-loop control algorithms. Martinez Paz et al. [141] addressed this issue by integrating signals from ORP and temperature sensors, enabling an intelligent flushing function. Compared to a fixed five-minute flushing strategy, this approach can reduce flushing time and water consumption by up to 46%. The wireless sensor system is illustrated in Figure 3.

In summary, ORP can be applied in real-time monitoring of disinfectant residuals, optimizing flushing strategies, and controlling chlorination processes, thereby playing a crucial role in water-quality safety alerts and automated management systems.

4.2.4. DO

DO is a key parameter for assessing water quality and the health of aquatic ecosystems. Its accurate monitoring is crucial for environmental protection, aquaculture and drinking water safety [142]. The determination of DO can be divided into the Winkler method, electrochemical analysis method and fluorescence spectroscopy method, among which the Winkler method is not suitable for online monitoring [143]. The electrochemical analysis method using the Clark electrode directly measures DO through electrochemical reactions. Clark sensors have been used for decades, and the technique is sufficient against the limitations of calibration drift caused by oxygen consumption, flow dependence, electrical interference, and so on [144]. Fluorescence technology is mainly achieved through the quenching principle of fluorescence. It offers advantages such as fast response, lack of oxygen consumption, strong anti-interference capability and low maintenance requirements, making it a current focus of research and application [145].

The research direction in the area of DO detection seeks to integrate advanced microelectronics technology to ensure the accuracy and stability of sensors in long-term and online detection, while achieving automation, intelligence and multifunctionality in the analysis process [143]. To enhance the accuracy, stability and intelligence of DO monitoring, a variety of advanced sensing technologies and algorithms have been introduced into practical applications. In terms of sensing technology, nanomaterials and microstructures are widely employed to improve sensor performance. For instance, a study utilized nano-porous anodic aluminum oxide membranes to load photosensitive indicator platinum(II) porphyrin, constructing a highly sensitive optical sensor. Its response intensity reached 26.2, much higher than the 6.5 observed with conventional substrates, significantly improving detection resolution and sensitivity, and enabling the identification of subtle changes in DO [142]. On the other hand, flexible sensor technology has also made significant progress. For example, a dual-electrode sensor encapsulated in polydimethylsiloxane was integrated into a bionic robot fish, achieving three-dimensional dynamic monitoring of dissolved oxygen in aquaculture waters, along with an error of less than 0.2 mg/L, demonstrating excellent stability and applicability [146].

At the algorithmic level, machine learning and deep learning are widely employed to enhance the intelligence of monitoring systems. For instance, Wang et al. [147] utilized transfer learning and convolutional neural networks to process river image data from surveillance cameras. By extracting visual features such as water color, suspended matter and flow state, they achieved high-precision prediction of DO, providing a new approach for low-cost and large-scale water-quality monitoring. Additionally, the process-guided deep learning model was applied to predict the daily extreme and average DO concentrations at different river sites. Although it did not significantly outperform traditional methods, it demonstrated the potential of integrating process mechanisms with data-driven approaches. Shaghaghi et al. [148] developed an innovative system based on the IoT platform, integrating pulse-oximeter sensors and machine learning algorithms. This system provides a low-cost and sustainable real-time DO monitoring solution for aquariums, aquaculture and other fields. The core objective of this system is to address the drawbacks of traditional dissolved-oxygen measurement devices, such as high cost, inefficiency and reliance on manual operation, while achieving economic, efficient and automated water-quality management.

In conclusion, DO monitoring is advancing towards multi-technology integration, increased intelligence and systematization. The development of new sensitive materials and sensors has significantly enhanced the accuracy and stability of monitoring, while machine learning and deep learning algorithms have greatly improved data analysis and prediction capabilities. Together, these advancements are driving the reliable application of DO monitoring in fields such as field investigations, aquaculture, drinking water safety and environmental emergencies.

4.2.5. Turbidity

Turbidity is a measure of the relative clarity of water, as suspended water particles impart its visual properties. It is measured based on the loss of clarity associated with particles in water; these may cause light absorption, reflection, or dispersion [149]. A scattering-type turbidity sensor is used to measure the scattering of light by suspended particles due to scattering phenomena. Sensors emit light as sources, measure the scattered light at different angles and infer the turbidity of liquids by analyzing the intensity of the scattered light. A transmissive turbidity sensor infers turbidity by measuring the degree of light transmission in a liquid. The light source is located on the side of the liquid, and the sensor measures the intensity of light passing through the liquid. Transmission-type sensors are commonly used for the turbidity measurement of transparent liquids. With the increasing automation and intelligence of detection technology, online turbidity analyzers combine the sensor types used for scattering and transmission. By simultaneously measuring the intensity of scattered and transmitted light, turbidity can be determined based on the ratio of scattered and transmitted light or the sum of the two, providing more accurate turbidity information [150]. The selection of the appropriate type of turbidity sensor depends on the specific application requirements, such as the measured properties of the liquid, particle size and concentration range.

In the development of the turbidity monitoring algorithms, signal processing and intelligent modeling played a core role. For example, to enhance the signal-to-noise ratio under low turbidity conditions, Tang et al. [151] developed a sensor system based on the principle of orthogonal demodulation, using an 860 nm near-infrared LED as the light source. Through modulation and demodulation technology, they extracted weak photoelectrical signals and achieved high-precision measurements within the range of 0–5 NTU, with a relative error of ≤±1% and a detection limit as low as 0.0049 NTU, fully meeting the demanding requirements of drinking water quality monitoring. On the other hand, machine learning algorithms have been successfully applied to turbidity classification and prediction. Parra et al. [152] collected data using a RGB multispectral light source combined with 64 light combinations, and achieved turbidity quantitative prediction with an R² of 0.979 by employing principal component analysis in combination with Gaussian process regression. They also used a K-nearest neighbor classifier to achieve a turbidity grade classification accuracy of 91.23%.

In summary, monitoring water turbidity through sensors has transformed from traditional single optical measurement methods into a comprehensive technology that integrates advanced optical design, intelligent algorithms and IoT systems. This evolution not only provides accurate and reliable data but also plays an indispensable role in smart water management, environmental protection and public health.

4.2.6. Electrical Conductivity (EC) and Salinity

EC refers to the ability of water to carry charges when it contains minerals and salts, while salinity measures the amount of the dissolved salts in the water [153]. The minerals and salt ions most commonly found in water are Na⁺, Mg²⁺ and K⁺. The higher the concentration of salt ions, the higher the EC value. Therefore, EC sensors are widely used to estimate salinity, with temperature’s effects on conductivity often needing to be accounted for. By employing various algorithms, particularly machine learning methods based on sensor data, the accuracy and real-time performance of salinity monitoring can be further enhanced [154].

In addition to EC sensors, refractive index sensors and chemical sensors are also used for salinity measurement. For example, refractive index sensors indirectly estimate salinity by measuring the refractive index of light in water, as changes in salinity affect the refractive index [155]. In practice, when using optical sensors for refractive index measurements, algorithms such as support vector machine regression and linear regression can be applied to improve the accuracy of salinity estimation [154,156]. Meanwhile, chemical sensors measure the concentration of specific ions through an ion-selective electrode, and the potential change of the electrode directly reflects changes in salinity [156,157,158]. These sensors are widely applied in fields such as agricultural water-quality monitoring and river basin management.

To further enhance the effectiveness of salinity monitoring, researchers have integrated various sensor technologies. For example, in agriculture, the use of nickel–hydrogen bridge metal–organic framework sensor arrays for real-time monitoring allows accurate measurement of K⁺, NO₃⁻ and pH values. Its excellent stability and high sensitivity make it particularly suitable for use in harsh environments and ensure continuous water-quality monitoring [159]. Through wireless data transmission and portable devices, these sensors can provide real-time water-quality data in remote areas. RS technology also offers a novel perspective for water-quality monitoring via sensors. The effective combination of the two has successfully simulated water-quality parameters such as DO and EC, with data modeling being conducted using the support vector regression algorithm. In a study of two small rivers in the Thracian region of Greece, combining multi-season and multi-sensor data resulted in a high R² value, demonstrating that integrating RS and sensor data can significantly enhance the accuracy and reliability of water-quality monitoring [154]. Additionally, the use of graphene-based surface plasmon resonance sensors for salinity measurement has shown high sensitivity and wide detection accuracy, making this technology a powerful tool for efficiently monitoring water salinity concentrations [156].

Through sensors and advanced algorithms, modern water-quality monitoring is capable of dealing with complex water environments and rapidly changing situations, and providing effective data support and decision-making basis for various fields.

4.2.7. Total Dissolved Solids (TDS)

The total amount of dissolved organic and inorganic elements in water is represented by TDS [160]. According to their working principle, TDS sensors come in various types. EC is an indicator of TDS in water, so TDS can be estimated by measuring the EC of water. Akram et al. [161] developed an EC and TDS sensor and confirmed a linear positive correlation between TDS concentration and EC voltage value. Electrochemical sensors use electrodes to react with ions in water to measure TDS. Typically, the potential change of the electrode is related to the concentration of TDS in water. Optical sensors use optical technology to measure the concentration of dissolved solids in water. This may include techniques such as light absorption, light scattering, or fluorescence, in which optical sensors can provide highly selective measurements for certain specific ions or molecules. Feng et al. [162] used manufactured fiber optic sensors for in situ application of TDS, demonstrating performance levels comparable to the EC method. Chemical sensors use specific chemical reactions to detect dissolved solids in water. They use specific reagents or reactions to infer TDS concentration by measuring the properties of reaction products. Some TDS sensors use changes in vibrating electrodes to measure the concentration of TDS in water. When the concentration of TDS in water changes, the vibration frequency or amplitude of the electrode will also change. Microfluidic sensors utilize microfluidic technology to measure TDS through the diffusion of dissolved solids in microchannels, providing fast response and high sensitivity [163]. Each type of TDS sensor has its unique advantages and applicable scenarios. The selection of the appropriate sensor type usually depends on specific application requirements, such as accuracy requirements, measurement range, selectivity for specific soluble solids, and so on.

Currently, the core development trend of TDS sensor monitoring is shifting towards a distributed intelligent architecture. In the future, by leveraging advanced models such as federated deep learning [164], it will be possible to achieve high-precision monitoring while ensuring robust data privacy and security. Through the fusion of multiple sensor parameters, the system can significantly enhance its anti-interference ability and reliability and greatly reduce reliance on network transmission and centralized computing. This technological approach significantly improves the feasibility of deployment in remote and resource-constrained areas, driving water-quality monitoring towards real-time, scalable and secure development.

4.3. Practical Applications of Sensors

Sensors have been employed in the monitoring of surface water and wastewater. The applications of sensor water-quality monitoring are shown in

Table 3. Applications of sensor techniques to water-quality monitoring.

Area	Sensor	Monitoring Index	Algorithm	References
Dormitory secondary water supply system	Raspberry Pi-based multi-sensor system	pH, TDS, temperature, turbidity	RF	[165]
Irrigation systems	Portable multi-sensor device	pH, K+, NO3−	Nernst equation	[159]
Drinking water	Online turbidity monitoring module	Turbidity	Scattering/reference specific value model, Scattering/transmitting specific value model	[166]
Environmental water bodies	Online turbidity monitoring module	Turbidity	Transmitting/reference specific value model	[166]
In situ water reuse system	Online sensors	ORP, pH, EC, temperature, turbidity	Logistic regression	[167]
Urban sewage treatment plant	Integrated screen-printed electrochemical sensor	DO	_	[168]
Salt water	New surface plasmon resonance biosensor	Salinity	Transfer matrix method	[156]
The Qazi Ahmed town canal in Sindh, Pakistan	IoT sensors	TDS, EC	Federated deep learning	[164]

.

4.3.1. Natural Water-Quality Monitoring

Sensors play an important role in monitoring the water quality of natural water bodies, providing real-time, accurate and comprehensive water-quality data. They can measure indicators such as the pH, conductivity, DO, turbidity and temperature of water bodies, factors which have important indicative significance for the ecological health and water quality of water bodies. The optofluidic monitor proposed by Cheng et al. [169] incorporated various electronic sensors to improve the automation and detection efficiency of online water-quality analysis. Compared with the standard methods of that nation, the relative measurement error was only about 6%, and the detection limit was 0.05 mg/L. The technique meets the measurement requirements for most surface waters and is an effective online automatic monitoring technology for water quality. Gao et al. [165] developed a multi-sensor water-quality monitoring system based on a Raspberry Pi. Using machine learning algorithms, they achieved efficient, real-time monitoring of water-quality parameters such as pH, TDS, temperature and turbidity. They combined these parameters to obtain the overall water-quality situation. The RF model demonstrated an accuracy of 98.1%, offering reliable technical support for water supply safety and intelligent early warning systems. Abdollahzadeh et al. [159] developed a portable multi-parameter sensing device based on nickel–hydrogen bridge metal–organic framework technology that can simultaneously monitor the K⁺, NO₃⁻ and pH values in agricultural water. This system employs an innovative solid-state electrochemical sensing structure, ensuring excellent stability. The device integrates wireless transmission functionality, achieving continuous, precise monitoring for up to 20 h in real-world agricultural irrigation environments, unaffected by pH fluctuations, and providing a reliable solution for intelligent agricultural water management. Chu et al. [166] developed a modular sensor water-quality monitoring device that utilizes both visible light and near-infrared light sources. By simultaneously measuring scattered light, transmitted light and reference light, and in combination with intelligent algorithms, it achieves precise monitoring of wide-range turbidity. The device exhibited a measurement error of less than 19.6% for low-turbidity tap water samples and less than 2.3% for high-turbidity environmental water samples, showcasing excellent anti-interference capabilities and strong commercial application potential.

A large dataset of sensor networks also provides a detailed spatiotemporal distribution of water quality data, thereby allowing researchers to better understand the functions and change mechanisms associated with water ecosystems. For example, Luo et al. [170] established an online water-quality warning system for Qiandao Lake based on a sensor based high-frequency monitoring system. The collected data could help water environment managers identify and predict the impacts of extreme climate events, issue early warning signals, and provide key water-quality information for drinking water transported to Hangzhou City. The application of sensors in monitoring the water quality of natural water bodies has promoted the modernization of water environment monitoring, provided high spatiotemporal resolution water-quality data, and offered strong tools for scientific research, environmental protection and water resource management to comprehensively and in real time understand the dynamic changes in water ecosystems. Sensors are increasingly being integrated with machine learning to provide more accurate monitoring data. Das et al. [164] proposed a federated deep learning framework to monitor TDS and EC, addressing the scalability challenges and noise sensitivity of traditional machine learning models. The results demonstrated that this model performed exceptionally well, achieving an accuracy rate of 98%. It also offered advantages such as data privacy protection, low-power computing and strong scalability, making it particularly suitable for resource-constrained environments. Federated deep learning can be integrated into low-cost devices, adapting to the needs of different regions. In the future, it has the potential to further enhance data reliability and expand its application scope.

By strategically integrating artificial intelligence and machine learning technologies with sensors, the system will be able to understand and adapt to subtle variations in water quality, making improvements based on long-term monitoring data [23].

4.3.2. Water-Quality Monitoring in Wastewater Treatment

Sensors play a crucial role in wastewater treatment, providing support for achieving efficient, intelligent and automated wastewater treatment. Sensors are used in wastewater treatment to monitor and analyze key water-quality parameters and environmental data in real-time, playing a crucial role in controlling biological treatment processes, optimizing chemical reaction rates, and ensuring system operation under appropriate conditions.

By integrating sensors with automatic control systems, real-time monitoring and intelligent control of the processing process can be achieved. Sensors monitor and process the status and performance-based parameters of equipment to achieve real-time monitoring of its condition. Bo et al. [171] designed an IoT online monitoring system for mine water and developed a wireless monitoring experimental platform for the same. The polluted mine water caused by underground mining activities in coal mines needed to be treated before discharge; otherwise, it would pollute the environment and waste resources. The platform could collect, transmit and store real-time information on the quantity and quality of mine water; quickly and accurately reflect the dynamic changes in the nature of mine water; and detect most abnormal data associated with mine water, meeting the requirements of timely detection in mining water treatment and utilization scenarios.

The large amount of data generated by sensors is used to establish historical records, perform trend analysis and generate reports. This is crucial for evaluating system performance, developing future operation and maintenance strategies and meeting regulatory and environmental regulations. For example, Reynaert et al. [167] developed a real-time water-quality monitoring system based on machine learning algorithms. This system integrated low-cost and widely available OPR sensors, achieving accurate predictions of water-quality safety by maintaining a false alarm rate within 2%. This innovative solution requires only 1–2 core sensors for efficient monitoring, offering a practical, cost-effective and reliable solution for water reuse applications. In fact, the use of the sensors extended beyond monitoring the aforementioned water-quality parameters. In specific scenarios, such as those involving effluent from sewage treatment plants, the demand for monitoring nutrients is particularly prominent. Simply tracking conventional parameters is no longer sufficient; sensors must be integrated into an IoT framework to enable remote monitoring of additional parameters like nitrate and ammonium salts [172]. Notably, most sensors currently only measure a limited number of parameters such as DO and pH, while the monitoring of key nutrients like TN and TP still depends on indirect calculations or multi-source data fusion [170]. Sensors can also be used for BOD analysis. For this purpose, DO sensors must be miniaturized and capable of online monitoring to facilitate mobile water-quality assessments in environments such as reservoirs or farms. To address this challenge, Yamashita et al. [173] developed an open-type anode DO sensor, eliminating the need for the OPM device required by traditional Clark-type sensors. This sensor can be directly used for on-site BOD monitoring. It is easy to install in aeration tanks and natural water bodies without protection or pretreatment. Furthermore, it can automatically clean electrode contaminants via a bottom aeration system, significantly extending its service life. To further enhance portability, Duan et al. [168] proposed a micro DO sensor that adopted a printed three-electrode microchip and a paper filter disk structure. The copper/carbon nanomaterial-modified electrode functions as the working electrode, supporting low-power device connectivity and enabling real-time on-site analysis via mobile phones.

Overall, the widespread application of sensors in sewage treatment provides strong support for improving treatment efficiency, reducing operating costs and ensuring environmental safety, promoting the development of the sewage treatment field towards intelligence, efficiency and sustainability. In the future, water-quality monitoring using sensors will also require support from solar energy and mobile power banks to ensure continuous data collection without energy limitations.

4.4. Benefits and Limitations of Sensors

4.4.1. Benefits of Sensors

Sensors have multiple advantages in remote water-quality monitoring, making them an effective tool that supports real-time, accurate and continuous water-quality monitoring [139].

Firstly, intelligent sensors can process and analyze the collected data in real time, and communicate with other devices in real time through wireless or wired means. They can transmit monitoring data to multiple platforms in real time through network technology, and share information with and communicate with multiple parties. Secondly, by using wireless communication technology to connect a large number of intelligent sensor nodes, a sensor network with a wide coverage area can be formed, a network which can facilitate data exchange and collaboration, achieve distributed monitoring and control and have high flexibility and scalability. Thirdly, the automatic calibration technology of intelligent sensors can reduce the error of the sensor itself, thereby improving the accuracy of monitoring results. In addition, intelligent sensors can analyze data in real time based on preset rules and warning values by using efficient data analysis algorithms and models. Once an abnormal situation is detected, the internal integrated warning system can trigger corresponding warnings and emergency response measures.

4.4.2. Limitations of Sensors

Although sensors play an important role in remote water-quality monitoring, there are also some limitations and challenges that need to be considered in practical applications.

Due to the different properties of water-quality parameters, it is difficult to integrate all different parameter detection techniques into a single system [174], and it is necessary to select suitable water-quality parameters and sensors. Moreover, the stability and accuracy of sensors in remote environments may be affected by environmental factors such as temperature and humidity, which can affect the accuracy of monitoring data and require regular calibration. Additionally, the large amount of data generated by sensors requires real-time transmission and storage, which may involve high communication and cloud storage costs. In addition, sensors require a stable power supply, especially for long-term remote monitoring. Places where sensors are used may not have convenient power sources, so it is necessary to consider using technologies such as low-power sensors, solar energy, or other renewable energy sources for power supply. Abnormal situations detected by sensors may require remote operation for processing, and it is necessary to ensure that the monitoring system has a certain degree of automation and remote-control capabilities.

4.4.3. Comparison of Sensors and RS Technology

In water-quality monitoring, RS technology and sensor technology are two fundamental approaches that, while distinct, are closely related. The primary distinction between them lies in their underlying principles and methods of data acquisition. Sensors are contact-based instruments that require direct immersion in the water body. They rely on physical or chemical reactions to accurately measure various parameters, such as pH, DO and EC, providing continuous, high-precision data at specific locations. These measurements serve as benchmarks for water-quality monitoring. In contrast, RS technology is a non-contact, macro-scale monitoring tool that uses sensors on aircraft, satellites, or UAV to collect electromagnetic wave signals reflected or emitted by the water body. Through inversion algorithms, RS technology estimates parameters related to the optical properties of water, such as Chl-a, TSS and CDOM. The main advantage of RS lies in its ability to quickly gather spatially continuous data over large water bodies with high efficiency. However, it generally offers lower accuracy and is limited to a smaller set of monitoring parameters compared to sensors.

Despite their differences, RS technology and sensor technology are deeply interconnected and complement each other. First, RS technology heavily relies on sensor data for calibration and verification. The development of RS inversion models and the assessments of their accuracy depend on synchronized and precise “ground truth” data, which is typically collected by on-site sensors. Without such sensor data, the reliability of RS inversion results would be compromised. Second, in practice, the two technologies form a highly efficient collaborative workflow. RS technology excels in large-scale scanning and rapid surveys of water bodies, identifying problem areas such as algal blooms and pollution plumes. Following this, sensor technology provides precise location tracking and in-depth analysis at specific points to confirm the issues and gather accurate quantitative data. Together, these technologies constitute the backbone of the “space–ground integrated” water-quality monitoring system. By combining the spatial trends captured by RS with the real-time, precise data provided by sensors, they offer a comprehensive and scientifically grounded foundation for decision-making in water resource management, pollution control and ecological research.

5. Conclusions

Water-quality monitoring plays a vital role in protecting and managing water resources, maintaining ecological balance and safeguarding human health. Remote data acquisition and treatment methods, including RS and sensors, have been widely used for their intelligence and high accuracy.

These remote data acquisition methods can be used to measure a large number of parameters. The indicators including Chl-a, turbidity and TSM/TSS, CDOM, EC, DO, temperature and the value of pH were considered in analyzing the water quality with the help of RS technology, among which Chl-a, turbidity and CDOM can be analyzed directly. In sensor-based monitoring, the parameters of pH value, temperature, oxidation reduction potential, DO, turbidity, EC and salinity, and total dissolved solids could be determined using physical, chemical and biological sensors.

The remote data acquisition methods offer the advantages of being portable and efficient and demonstrating wide coverage, high resolution, adaptability and the ability to monitor in real time. However, this approach has some limitations. RS-based monitoring technology at the present stage still has limitations, such as difficulties in data analysis, large uncertainties and the inability to replace field monitoring. Sensor monitoring has been limited by the research and development in sensor production, calibration and maintenance requirements of sensors, cost of data transmission and storage and difficulty of remote intervention. Low-cost sensors are a supplementary alternative that can quickly detect changes in water-quality parameters. Further research is needed to develop operative and maintenance guidelines for these low-cost sensors.

Therefore, future researchers can choose the most appropriate method and data to determine the water quality, based on the available in situ data, the cost, and the geographical characteristics of the study area. In addition, the development of sensors, the improvement of data analysis and the update of remote intervention technology should be paid attention to in future research. By continuously optimizing the collaborative mechanisms existing between these two technologies and integrating emerging fields such as artificial intelligence and big data analysis, future water-quality monitoring will become more efficient and accurate, and capable of addressing increasingly complex environmental challenges.