Advances, Challenges, and Future Perspectives in Surface Water Quality Monitoring Using Remote Sensing and GIS: A Structured Literature Review

1  Introduction

Access to reliable freshwater supplies strengthens human health, underpins economic development, and supports the sustainability of natural ecosystems [1,2]. However, as global populations continue to rise and urbanization accelerates, pressures on freshwater systems intensify, leading to an increasing degradation of water quality. This decline poses serious risks to public health and threatens the ecological stability [3,4]. As the quality and availability of freshwater resources have become a pressing issue globally [5,6], water quality monitoring is necessary, particularly in water resource management. Timely identification of emerging problems through the trend analysis in water quality provides fundamental information for advanced management strategies [7,8]. To date, there has been a significant water quality assessment and prediction, which enables the understanding of spatiotemporal water quality dynamics and provides scientific support for aquatic environmental protection.

Water quality monitoring is traditionally done manually, collecting water samples from various sources of water and analyzing them under controlled laboratory conditions. The samples are analyzed for their physicochemical and biological parameters that affect the water quality [9]. However, given that this method is considered accurate, it is time-intensive and requires human expertise, equipped laboratory facilities, and a considerable amount of investment during extensive field work for data collection. Additionally, due to the time required for field sampling and laboratory analysis, it is not possible to get immediate results to respond to water pollution emergencies, thereby increasing the risk of environmental hazards [10]. Some of the traditional water quality monitoring approaches are: (1) on-site sample collection with lab analysis; (2) rapid on-site testing; and (3) automatic monitoring stations [11].

Although continuous monitoring of water sources in real-time water quality assessment offers numerous advantages, it is also impractical with the traditional methods [12], as a standalone solution for large-scale or long-term use. Some of the drawbacks of continuous water quality monitoring in real time are the initial cost of installation and maintenance due to the substantial price of specialized sensors, often calibration and cleaning for accurate and reliable data, and challenging interpretation due to the complexity of data management and analysis. From the challenges and limitations of using traditional water quality monitoring approaches, advances and innovations are highly significant, such as satellite-based remote sensing, machine-learning and data-driven models, integration of remote sensing with GIS, and complementary use of in-situ sensor networks and telemetry.

Remote sensing technology appears as a promising solution, which offers a time-efficient and large-scale monitoring capacity [13] and an effective approach to efforts and cost reduction [8] that complement traditional methods. It is also a solution in providing data that can be used for environmental management, research, and policy decisions.

Moderate to coarse spatial resolution of remotely sensed data from satellite imagery has been effectively shown to produce promising results in monitoring water quality in inland water bodies [14–16]. More recently, Unmanned Aerial Vehicles (UAVs) RS has drawn growing interest in scientific researchers and industry, due to its broad application predictions, such as in agriculture, forestry, mining, and other industries. UAVs can be flexibly provided with various sensors, such as optical, infrared, and Light Detection and Ranging (LiDAR), and have become an integral RS observation platform [17]. When a UAV is equipped with sensors with RS capabilities, it has the flexibility to collect information against exclusive reference while subject to remote human control. On the other hand, when the RS sensors were enhanced, the performance of the remote-control technology also improved. RS equipment compatibility was increased, the RS data reception and processing capabilities were developed, and UAV platforms and sensors became integrated, which gradually led to the development of RS applications with new capabilities [18].

One of the new capability of RS technology are Machine Learning (ML) techniques, which are capable of identifying the incremental patterns and trends that may be overlooked in traditional methods and making it remarkably advantageous in handling complex RS data [19,20]. Notwithstanding these advantages, there are still challenges in selecting the most effective ML algorithms for monitoring models and ensuring the reproducibility of analysis [21]. With the rapid advancement in the field of RS and the increasing availability of satellite data in the open domain, researchers are leveraging this data for water quality monitoring.

Another platform that incorporates spatial data, environmental variables, and RS outputs is the use of GIS. This is a broadly used tool for evaluating water quality and quantity. For example, the application of GIS techniques and Water Quality Index (WQI) in the assessment of water quality in the context of reservoir systems revealed to be a valuable contribution statistically to the understanding and management of connected water bodies [22], evaluation of surface water quality within a GIS framework by characterizing the physio-chemical characteristics of surface water from various locations [23]. GIS has developed rapidly to become efficient computing tool for various applications such as sophisticated 3D analyses and modeling [20], mapping the distribution of water quality parameters for identification of contamination hotspots [24], and predicting WQI values in unsampled areas to enhance the spatial coverage and enabling proactive decision-making in regions where direct water quality measurements are limited [25]. The WQI is a widely adopted tool for synthesizing various physical, chemical, and biological parameters into a single value that facilitates interpretation and decision-making [26]. When combined with GIS, WQI allows a spatial and temporal mapping of water quality and improved environmental assessments using methodologies such as statistical modeling and ML [27]. For instance, WQI was integrated with the GIS to spatially visualize and examine water quality data to identify the pollution hotspots, trend analysis for effective water resource management [28]. In response, the advances in RS and GIS technology offer an opportunity to demonstrate a better understanding to contribute to watershed management, pollution source identification, and environmental planning.

This paper aims to identify the critical gap by synthesizing recent advances in surface water quality monitoring using RS and GIS techniques and identifying future research opportunities and implementation challenges. This review compares the traditional water quality monitoring approaches with emerging RS data and GIS as a decision support tool. It further covers the fundamentals of RS for water quality, types of sensors, water quality parameters, retrieval algorithms, and analytical methods. Finally, this review consolidates the advances, challenges, and limitations to inform the development of current operational monitoring systems that support transparent, adaptive, and knowledge-based water resource decision-making.

2  Methods

This review was conducted in accordance with the PRISMA [29] guidelines to ensure transparency, reproducibility, and methodological rigor at each stage of the review process. The PRISMA checklist informed the procedures for study identification, screening, eligibility, and final inclusion (shown in Fig. 1). This structured approach helped reduce potential sources of bias and enabled reliable synthesis.

Figure 1: The diagram presents the structured literature review methodology used to identify and select studies on the advances, challenges, and future perspectives in surface water quality monitoring using RS and GIS.

2.1 Identification

A systematic search was conducted across three major scholarly databases, namely ScienceDirect, Scopus, and Web of Science (WoS), to identify studies relevant to this review. These databases were chosen to cover a wide range of scientific publications. This allowed comprehensive retrieval of relevant studies. The search included only articles published from 2012 through 2025. The research strategy employed a rigorously designed Boolean expression: (“water quality monitoring” AND “remote sensing”), (“water quality monitoring” AND “GIS”), (“water quality monitoring” AND “remote sensing” OR “machine learning”), (“surface water” AND “remote sensing” AND “GIS”) were used to refine the search, and duplicate records were removed using a reference manager.

2.2 Screening

In the screening stage, records were assessed based on titles and abstracts using explicit eligibility criteria. A total of 287 records were retrieved from ScienceDirect, Scopus, and Web of Science (WoS) and other sources (i.e., official websites and conference papers). After removing 29 duplicates, 258 records remained for screening. Of these, 105 records were excluded after title and abstract screening. 153 articles underwent full-text assessment, and 50 were removed because they were irrelevant or did not meet the eligibility criteria such as, (i) studies focused on domains outside the review scope (groundwater monitoring systems, land-only environmental monitoring, and marine ecological assessments without water quality parameter retrieval), (ii) absence of empirical application, and (iii) editorials or opinion papers. Ultimately, 103 studies were included in the final synthesis.

2.2.1 Data Extraction

The study was conducted using a standardized data collection form designed to obtain the most relevant information from each included study. The extracted data included key attributes (authors, year, study area), methodology, and major findings. To ensure accuracy and consistency, data from each included study were extracted individually by two reviewers, with no automation tools employed in the data extraction process.

2.2.2 Data Items

The initial outcomes of this review were the water quality parameters retrieved using RS and GIS techniques, and the methodological approaches used for estimation. Specifically, the methodological outcomes were extracted from the types of remote sensing platforms and sensors, retrieval algorithms, validations, and model performance. All relevant results for each, including study on the initial outcomes were collected.

In addition to the initial outcomes, other variables were extracted to maintain the contextual and methodological interpretation of the included studies. These include publication details, geographical setting, type of water body, and classification of study regions. There were no assumptions made about missing or unclear data, and only explicitly reported results were included in the synthesis.

2.3 Eligibility

The inclusion criteria for this review required that studies explicitly applied or discussed RS and GIS within the context of water quality monitoring, covering domains such as surface water quality monitoring and assessment, and applications related to water quality parameter retrieval. Only peer-reviewed journals and conference proceedings articles written in English and presenting clear methodological frameworks, validation strategies, or empirical findings were considered eligible. To maintain academic rigor and ensure the reliability of synthesis, unpublished sources were excluded from the analysis. Duplicate records identified across databases were removed using reference management software, followed by manual verification.

2.4 Final Inclusion

Following the full screening and eligibility assessment, a total of 103 studies met all inclusion criteria and were included in the final synthesis. These studies represent the most relevant and methodologically robust peer-reviewed evidence on the application of remote sensing and GIS in water quality monitoring within the defined scope of this review. This final inclusion ensured that the review was based on studies with clear methodological transparency and direct relevance to water quality monitoring and assessment. Additional references were retained to support background information, methodological explanations, and interpretation of findings.

2.5 Effect Measures

There is no quantitative effect measure used in this review since the findings were synthesized qualitatively. Instead, study findings were presented descriptively, which focused on the methodological approaches, retrieved water quality parameters, and reported model performance.

2.6 Synthesis Process

Studies were grouped for synthesis based on thematic and methodological characteristics, including: (i) traditional monitoring approach vs. RS technologies, (ii) remote sensing systems and observation platforms, (iii) water quality parameter types, (iv) retrieval methods, and (v) GIS Integration in water quality monitoring. Extracted data were classified according to the parameters and methods into consistent categories. Missing data were not cited, and only clearly reported results were included.

A qualitative narrative synthesis was conducted due to methodological heterogeneity among studies, and the results were presented using narrative summary tables with the key study characteristics, identifying patterns, methodological trends, strengths, limitations, and research gaps across studies. Variability was explored through subgroup analysis based on sensors, platforms, water bodies, and methods. No formal sensitivity analysis was conducted due to the qualitative nature of the synthesis and the absence of comparable quantitative effect measures. However, a greater weight was given to studies with strong validation strategies and larger datasets to support reliable interpretation of findings.

2.7 Risk of Bias Assessment

To ensure the robustness and credibility of the synthesized evidence, a qualitative risk of bias assessment was conducted for all included studies. Due to the methodological diversity of RS and GIS-based water quality research, a framework was developed based on commonly reported methodological limitations in the literature. Each included study was independently evaluated by two reviewers, and no automation tools were used in the risk of bias assessment process.

The assessment considered five key criteria: (i) quality of input data; (ii) validation strategies; (iii) model transparency and reproducibility; (iv) confounding factors; and (v) generalizability of results. Each study was qualitatively evaluated based on criteria and categorized as having low, moderate, or high risk of bias. This assessment was conducted during the data extraction phase and was used to inform the interpretation of findings, particularly in identifying methodological limitations and research gaps across the reviewed literature.

3  Results

To provide a general overview of the articles reviewed, Table S1 presents a summary of reviewed studies according to key attributes such as authors, year, methodology, study region, and major findings. Most studies focus on satellite-based monitoring of optically active parameters, while an increasing number incorporate machine learning and multi-sensor integration.

This section includes the studies focusing on groundwater monitoring systems, land-only environmental assessments, or marine ecological analyses without clear water quality parameter retrieval, which were excluded, as they did not align with the scope of this review. Additionally, studies without sufficient empirical application, such as conceptual frameworks, editorials, and opinion-based articles, were excluded due to the absence of methodological validation or quantitative results.

The risk of bias assessment, shown in Table S2, indicates that studies with strong validation strategies, sufficient in situ data, and transparent methodologies provide higher confidence, whereas studies relying on limited datasets or site-specific calibration exhibit greater methodological limitations. Due to substantial heterogeneity in study designs, data sources, and analytical approaches, no quantitative meta-analysis was conducted, and the synthesis was performed qualitatively. The regional distribution of selected studies was categorized as developed and developing countries based on World Bank [30] income classifications, which is also provided in Table S3 and Fig. S1.

Meta-analysis was not conducted in this review due to significant heterogeneity among the included studies in terms of data sources, remote sensing platforms, water quality parameters, and analytical methods.

As a result, summary effect estimates, confidence intervals, and measures of statistical heterogeneity were not calculated. Instead, findings were synthesized using a qualitative narrative approach to identify patterns, trends, and methodological differences across studies.

3.1 Traditional Monitoring Approach vs. Remote Sensing Technologies

The reviewed literature reveals two primary monitoring approaches: traditional ground-based techniques or field sampling methods and RS-enabled spatial monitoring frameworks. Shifting from traditional monitoring to RS technology does not position itself as a basic technological replacement but as a strategic adaptation to the identified spatiotemporal limitations in traditional systems.

Across the synthesized literature, it was evident that the identified limitations of traditional monitoring are that although laboratory-based analyses and in-situ sampling provide high analytical precision, it is labor-intensive, costly, and operationally constrained [31,32]. Likewise, the spatiotemporal variability in surface water bodies such as rivers, lakes, and reservoirs shows constraints due to the sampling or data collection conducted in fixed locations and discontinuous time intervals [33,34], and the limited accessibility to remote sites further restricts their spatial coverage [35]. The synthesis indicates that these framework limitations often produce spatiotemporal gaps, during which occasional pollution events or localized degradation may remain undetected [16]. These recurring observations in the literature have driven increasing research interest in monitoring approaches capable of broader spatial coverage and higher temporal frequency [36]. In response to these identified gaps, RS has increasingly become a methodological direction within the reviewed studies. RS-based approaches are consistently highlighted for their ability to provide a broad range of repeatable observations across a larger scale without requiring physical site access [15,37]. It can track trends using satellites and aerial platforms for data collection, and offers a valuable perspective for investigating different region types and emphasizes that satellite and airborne platforms improve the illustration of spatial patterns and temporal dynamics in water quality parameters, particularly in systems where manual sampling is sparse [38].

The literature further illustrates the broad application of multispectral sensors, such as Landsat, Sentinel, MODIS, MERIS, etc., for inland and coastal water quality monitoring. Behind the relatively high revisit frequency and accuracy for optically active parameters, these platforms still show continual challenges, including optical complexity in inland waters, atmospheric correction uncertainty, and sensor resolution constraints [11]. Hence, while RS expands its monitoring scope, its application requires thorough algorithm selection, calibration, and validation.

The reviewed studies also report an increasing interest in active and near-proximal platforms such as solar illumination, LiDAR, and Synthetic Aperture Radar (SAR), which generate their own signals, enabling additional structural and hydrological information [39]. More recently, unmanned aerial vehicles (UAVs) and unmanned surface vehicles (USVs) have been increasingly implemented for high-resolution and local-scale monitoring [40,41]. It shows that UAV-based hyperspectral and multispectral imagery, with the integration of machine-learning models, yields a high retrieval accuracy for turbidity, TN, TP, and Chl-a in rivers and small inland water bodies [42–44]. However, it was also reported that UAV-derived models are often site-specific, constrained by limited spatial extent and temporal continuity, and are less effective for capturing rapid hydrological fluctuations at broader scales. Accordingly, existing studies join in suggesting that UAV–RS integration is most effective as a complementary element in monitoring frameworks rather than deployed as a standalone solution for large-scale or long-term applications.

The analyzed studies illustrate a clear methodological transition: from ground-based and labor-intensive monitoring toward multi-scale, sensor-integrated systems designed to improve spatiotemporal representativeness in surface water quality monitoring.

3.2 Remote Sensing in Water Quality Monitoring

RS System Categories and Observation Platforms

The RS systems applied in water quality monitoring can be systematically categorized according to energy sources and observation mode: (i) Passive RS systems that uses the sun as the source of light and therefore can operate only during the daytime [45]; (ii) active RS and Near-Real-Time Monitoring System that uses their own energy and can operate during both day and night; and (iii) system observation platforms. Table 1 presents a summary of active and passive remote sensors and platforms for surface water quality monitoring, and their strengths and limitations.

Passive RS System

The reviewed studies demonstrate that medium-resolution multispectral satellites play a central supporting role in inland and coastal water-quality monitoring. Optically active RS systems exploit visible–NIR bands and band ratios to directly retrieve optically active parameters, particularly Chl-a, turbidity, and TSS. Conversely, chemically and biologically optically inactive parameters such as nutrients (TP and TN), DO, BOD, COD, and salinity, are indirectly inferred or proxy-based, which are calibrated with in-situ data and, in many cases, refined using machine-learning (ML) algorithms. For instance, Sentinel-2 MSI was used to retrieve multiple water-quality parameters—DO, CODMn, NH3-N, TP, TN, and turbidity—in the Yangtze River Delta; one-dimensional regression on NIR bands achieved sufficiently high inversion accuracies to demonstrate the feasibility of routine concentration mapping from MSI reflectance [46]. Similarly, Landsat-8 OLI data were coupled with LASSO regression and spectral indices to predict temperature, TDS, pH, turbidity, Chl-a, DO, and blue-green algae in the Tigris River; the model provided a cost-effective alternative to dense field sampling and enabled computation of a WQI that revealed persistently poor water quality [47,48]. Several studies highlight the strength of band-combination regression for single parameters. For example, in Lake Tana, ordinary least-squares models using specific Landsat-8 bands explained ~87% of turbidity variance, demonstrating that simple linear models on carefully chosen bands can be robust for sediment-rich waters [49]. Likewise, empirical band-ratio models from Sentinel-2 (ratios such as B5/B4, B4/B3, and B4) effectively mapped spatiotemporal patterns of Chl-a, turbidity, and TSS in the Koka Reservoir, supporting diagnosis of eutrophication and shallow-water sediment dynamics [50]. Overall, medium-resolution multispectral satellites perform best in medium-to-large rivers, lakes, reservoirs, and sediment-rich water. It consistently permits repeatable retrieval of optically active water-quality parameters at regional scales. However, these satellite estimates on optically inactive parameters as proxy-based are sensitive to calibration density, AC, and water-type variability, which indicates the need for further strategies.

A large cluster of studies further shows that combining Landsat and Sentinel-2 with advanced machine-learning algorithms greatly improves retrieval accuracy and allows multi-parameter prediction. For instance, RF, XGBoost, support-vector regression, and deep neural networks applied to Landsat/Sentinel-2 reflectance achieved high R2 for turbidity [51], TSS [52,53], Chl-a [54,55], nutrients such as TN and TP [56–58], and algal bloom [59] in rivers, lakes, and coastal waters, often outperforming classical semi-analytical algorithms. In these works, the major strengths are high predictive power (often R2 > 0.8), ability to handle multi-parameter datasets, and explicit mapping of spatiotemporal pollution patterns such as eutrophication hot spots, urban plume zones, and seasonal turbidity peaks. ML-based multispectral retrieval works well for multi-parameter mapping in rivers, lakes, and reservoirs with sufficient in-situ training data by capturing nonlinear spectral relationships, but performance degrades outside calibration domains, highlighting the need for spatial–temporal cross-validation and transferable model design.

Finally, some studies leveraged very large archives of Landsat reflectance to model TSS, TN, TP, DO, and Chl-a across many rivers, revealing continental-scale patterns and demonstrating that neural-network models trained on long time series can robustly generalize across diverse hydrological conditions [52,60–63]. While the key strength of passive multispectral RS obtains multi-decadal, spatially continuous records of inland water quality, it is most effective when used for large scales, such as large lakes, major river basins, and coastal waters. This demonstrates the unique strength of passive RS in resolving long-term and large-scale water-quality trends. However, when a need arises for a short-term or local scale, higher-frequency observations are still needed.

A second major group uses ocean-color sensors and other coarse-resolution passive instruments to monitor regional to global patterns of phytoplankton biomass, CDOM, nutrients, and related biogeochemical indicators. For instance, MERIS and Sentinel-3 OLCI products, often processed with C2RCC, MPH, and quasi-analytical algorithms, were shown to retrieve Chl-a, Secchi depth, and turbidity with overall regressions close to 1:1 relative to in-situ data across >100 German lakes, confirming their value for national-scale monitoring [64–66]. Similarly, long-term products such as MYDOCGA-MODIS reflectance were used with XGBoost to map TN and TP in Hong Kong’s coastal waters, revealing both hot spots (e.g., Shenzhen Bay) and decadal declines in nutrient levels, thereby showcasing the strength of ocean-color time series for tracking long-term management outcomes [67]. Ocean-color sensors are best suited to open coastal waters and large lakes with medium to large scales. It substantially performs in homogeneous conditions but performs poorly on a smaller scale due to the coarse resolution. Therefore, these constraints required downscaling or complementary higher-resolution sensors for small inland waters. This is further proven when [68] downscaled the 20-m resolution SWIR band from Sentinel-2 images to 10 m based on pan-sharpening, which shows a more detailed spatial information of water bodies. Also, ref [69] highlights that comprehensive water data listing remains insufficient, especially for small water bodies, such as ponds, which are overlooked regardless of their ecological significance.

Other studies compared different processing chains and products—for example, JAXA G-Portal vs. JASMES GCOM-C/SGLI retrievals—showing that multi-product ensembles can reduce uncertainty in Chl-a and TSM estimates, while still providing globally consistent coverage [70–73]. In coastal and estuarine seas, products from MODIS, VIIRS, COCTS, and GOCI, combined with empirical, semi-analytical, and ML algorithms, proved particularly effective for harmful algal bloom (HAB) monitoring and deoxygenation assessments, for instance, detecting cyanobacterial blooms in Lake Erie and summer coastal deoxygenation associated with reclamation projects [71,74–77]. Some contributions extended this to global biogeochemical indicators, such as remote retrieval of sea-surface nitrate trends from MODIS, demonstrating decreasing SSN linked to global warming and ENSO variability [78], or to lake climate-product evaluation, where CLMS/ESA lake products showed robust LSWT retrievals and usable Chl-a and turbidity information at the continental scale [79]. The advantages of coarse-resolution passive missions have been evident in providing long, consistent, and often near-global coverage needed to understand climate-scale and basin-scale water-quality change.

A third cluster in the review literature emphasizes high-spatial or high-spectral-resolution passive sensors for local-scale, detailed water-quality assessment. For example, in the study of Di Francesco et al. [8], PlanetScope microsatellites and Sentinel-2 were shown to accurately retrieve Chl-a, turbidity, and cyanobacteria in small lakes using semi-empirical indices and ACOLITE processing, demonstrating the cost-effective potential of daily microsatellite constellations for near-real-time water-quality mapping. Coccia et al. and Lei et al. [80,81] at even finer scales, airborne hyperspectral sensors such as CASI-1500 and AMMIS enabled precise mapping of pond depth, turbidity, Chl-a, TN, and TP, distinguishing restoration success and identifying algal-bloom-affected shorelines that would be missed by coarser sensors. These high-resolution commercial satellites are effective for small lakes and urban rivers, and they enhance the detection of fine-scale optical variability. However, resolving the finer spatial scales of these satellites also became their limitations due to the limited coverage, cost, and operational constraints, which restrict deploying on a larger scale.

Proximal hyperspectral systems (HPSs) were also deployed over lake surfaces to retrieve ultra-high-frequency lake surface temperature and related thermal dynamics via ML models such as XGBoost, DNN, and KNN, providing an effective early-warning tool for harmful algal bloom risk and heatwave-driven stratification disruption [82]. The key strength of these high-resolution passive systems is therefore their ability to resolve fine-scale heterogeneity in shallow, small, or complex water bodies and to support intensive local management and restoration actions. However, it is limited to spatial representativeness, which necessitates integration with satellite or UAV observations for broader applicability.

A final set of passive-sensor studies in the reviewed studies emphasizes data fusion and multi-sensor integration to leverage complementary strengths. For example, an advanced spatiotemporal fusion model (ESTARFM) combined Landsat-8 OLI and MODIS to generate higher-frequency SPM maps with improved accuracy compared to alternative fusion methods, showcasing how fusion can yield both high spatial and temporal resolution for sediment monitoring [83]. Similarly, Sentinel-2 and Sentinel-3 imagery were integrated to derive harmonized Chl-a products and uncertainty fields via Bayesian neural networks and mixture density networks, enabling consistent cyanoHAB monitoring across sensor platforms and supporting public-health risk management [84]. Multi-mission constellations (e.g., GF-1, Landsat-8, Sentinel-2) combined through genetic-algorithm-optimized RF were also shown to give superior TN/TP retrievals compared with single-sensor models [43,85–87]. Using data fusion and multiple sensors performs best across heterogeneous waters and cloudy regions by combining the spatial and temporal strengths of different sensors, although it increases the methodological complexity.

Together, these studies highlight that passive techniques are strongest when they exploit multi-sensor synergies, providing: (i) higher revisit frequency, (ii) better handling of cloud gaps, and (iii) improved robustness of inversion models across varying optical regimes.

Active RS and Near-Real-Time Monitoring System

Throughout this review, the term active remote sensing is taken for radiometric systems, whereas in-situ telemetry and RTRM systems are discussed as near-real-time monitoring systems that complement, but do not replace, remote sensing observations.

Among the active systems, LiDAR-based techniques stand out for their ability to probe the vertical dimension of the water column. For instance, shipborne lidar using the Fernald method successfully retrieved Chl-a and optical properties in Qiandao Lake, with findings indicating that LiDAR has strong potential for long-term monitoring of inland lakes and that the proposed retrieval could be transferred to other sites [88]. In another study, marine LiDAR with laser-induced fluorescence was combined with Sentinel-2 MSI to estimate Chl-a, CDOM, and TSS. Regional bio-optical models derived from these active–passive combinations showed good applicability across both eutrophic and oligotrophic waters, demonstrating that LiDAR can sharpen satellite-derived bio-optical algorithms and extend them to optically diverse conditions [89]. These LiDAR applications illustrate the main strength of active optical systems. It provides vertically resolved, highly sensitive measurements of water constituents, which can be used both for direct monitoring and for calibrating passive satellite algorithms in complex waters. Yet, LiDAR performs best in optically complex lakes and reservoirs; there are still limitations that may be improved, such as high operational costs and limited coverage, which restrict its use to calibration, validation, and process studies.

Another set of active techniques involves satellite radar altimeters and associated microwave missions, often used in combination with optical imagers. For example, multi-mission radar altimetry (TOPEX/Poseidon, ERS-2, Envisat, CryoSat-2, SARAL) combined with Sentinel-3 OLCI observations were used to assess trends in lake water level, Chl-a, and turbidity. Boosted regression trees showed that increases in Chl-a were strongly linked to catchment built-up area and shrinking water surface, underscoring the strength of altimeter–ocean-color synergies for diagnosing land-use impacts and climate-driven hydrological change [90]. Microwave and thermal IR data were also used to infer DO and related variables at coastal scales; for instance, MODIS and VIIRS products, together with SST and Chl-a, allowed multiple regression models to detect significant summer deoxygenation associated with reclamation projects in the Yellow Sea [75], while other MODIS-based work retrieved global sea-surface nitrate trends [78]. These demonstrate that active microwave missions, though less frequently used than optical sensors, add unique information on water level, circulation, and large-scale nutrient dynamics that are difficult to obtain from passive visible–NIR data alone. Commonly, radar and microwave missions are most effective in large lakes, flood plains, and coastal waters for water level and surface dynamics; however, their weak sensitivity to water-quality constituents limits direct WQP retrieval, emphasizing their complementary role.

In addition, this review also includes real-time remote monitoring (RTRM) systems, which are essentially active in-situ sensor networks sending data remotely rather than remote-sensing in the classical radiometric sense. For instance, RTRM installations measuring DO, pH, conductivity, turbidity, and sediment concentration in marine waters were interpreted with fuzzy systems and neural networks to characterize the spatial vulnerability of coastal zones, identify areas where DO remained supportive of aquatic life, and highlight recreational disturbances from boating [91]. The strength of these systems lies in their high temporal resolution and direct measurement capability, which complements both passive and active satellite observations that are limited by overpass time and atmospheric conditions.

These RTRM and USV/UAV platforms can be viewed as active monitoring technologies that, when combined with satellite and airborne imagery, create multi-scale, multi-sensor observing systems capable of capturing everything from sub-hourly events to multi-decadal trends. For instance, the consistent success of Landsat and Sentinel-2/3 in retrieving Chl-a, turbidity, TSS, DO, and nutrients in very different water bodies—rivers [48,51,92], lakes and reservoirs [46,50,93–96], bays and coasts [85,87,97,98], and open sea/coastal systems [67,71]—is a central strength emphasized in the reviewed studies.

Active techniques (LiDAR, radar altimetry) and near-real time monitoring systems (RTRM networks), while fewer in number, add depth, structural information, and temporal continuity: LiDAR yields vertical profiles of Chl-a and optical properties, strengthening calibration/validation in optically complex lakes [88,89]. Radar altimetry and related missions capture water-level dynamics and large-scale nutrient/temperature changes, enriching interpretations of satellite-observed Chl-a and turbidity [78,90]. Passive optical sensors are the backbone of water-quality RS, while active techniques, near-real time monitoring systems, and unmanned platforms provide critical added strengths, vertical profiling, hydrological context, and real-time local detail, that together form an integrated multi-sensor observing system for surface-water monitoring. Near-real-time monitoring systems work best in managed rivers, lakes, and coastal zones, requiring rapid response by providing direct, high-frequency measurements. Yet, spatial representativeness is being limited due to its point-based nature and maintenance demands, which can be further looked for improvement.

Observation Platforms

Beyond sensor type and energy source, RS performance is also strongly governed by the observation platform on which sensors are deployed.

The reviewed literature also dedicates several entries to unmanned platforms (UAVs and USVs) carrying passive RGB or multispectral cameras, which bridge the gap between in-situ measurements and satellites. For instance, drone-based RGB imagery combined with ML models such as RF, GBM, XGB, and KNN accurately predicted E. coli concentrations in irrigation ponds, demonstrating that simple RGB indices, when collocated with water-quality samples, can provide high-resolution microbial risk maps for local irrigation management [99]. Similarly, UAV-borne multispectral cameras with indices like NDCI, NDAI, MCI, and CI were combined with satellite imagery and in-situ geophysical measurements to map stream-scale algal patches and guide buffer-zone and pollutant-load management [100]. Other UAV-based shows advanced ML architectures (e.g., GA-optimized XGBoost, TLNet, CNN, LSTM, Transformer) accurately retrieving TP, TN, DO, TSS, and Chl-a in small lakes and urban rivers, capturing sharp spatial gradients linked to aquaculture, urbanization, and point sources [43,101]. On the surface, USVs equipped with Chl-a sensors were used to generate spatially interpolated Chl-a maps using Kriging/IDW, showing that autonomous surface vehicles can efficiently collect high-density datasets over streams and reach-scale systems, and the resulting maps are promising for the commercialization of remote monitoring technologies. UAV-based monitoring is most effective in small lakes, narrow rivers, and urban water bodies, which significantly improve local-scale water-quality characterization and satellite validation through ultra-high spatial resolution. However, standardized protocols and satellite integration are needed because it is site-specific, and temporal continuity is limited. Moreover, USVs perform well in rivers and reservoirs, which enhances spatial sampling density and bridges gaps between point measurements and imagery. However, its reduced scalability and weather sensitivity indicate a need for improved calibration and interpolation rather than standalone monitoring.

The strength of these unmanned-platform approaches is flexibility and ultra-high spatial detail: they can be rapidly deployed over small or narrow water bodies where satellite pixels are too coarse, and they provide near-instant feedback for local managers, while still integrating seamlessly into satellite-based frameworks.

These findings reflect variability in monitoring performance across different approaches, which is influenced by spatial coverage, temporal resolution, and methodological approach, highlighting the heterogeneity of the reviewed studies.

3.3 Water Quality Parameter

Water quality is essential for ecosystems, human health, and economic development. Typically, it is defined by its suitability for a specific use, which is determined by optically active parameters (Chl-a, TSS, SDD, CDOM, turbidity, surface water temperature), which directly influence water-leaving radiance and optically inactive parameters (pH, TN, TP, DO, COD, BOD/CD, and salinity), which lack distinct spectral signatures and are therefore indirectly inferred through proxy relationships, auxiliary variables, and site-specific calibration. Across both optically active and inactive water quality parameters, retrieval performance is dependent on the transferability, confounding factors (AC and adjacency effects), and calibration and validation design. Transferability across regions and seasons remains limited unless models are trained on diverse hydro-optical conditions and evaluated using independent spatial–temporal validation schemes. Thus, the atmospheric correction, i.e., the removal of atmospheric effects, is an essential prerequisite for obtaining accurate watercolor information [102]. The adjacency effect is caused by reflected photons from surrounding land targets that are scattered by atmosphere components into the sensor’s field of view [103]. This complex phenomenon reduces the contrast between the high-reflectivity (land) and the low-reflectivity (water) surfaces. As a result, scattered photons from the land targets close to the water bodies can distort water spectral reflectance, mainly affecting small water bodies [104]. Additionally, the absence of in situ measurements for calibration and validation raises accuracy concerns; the models that have been developed using RS data need to be properly calibrated and validated using in situ data [105–107]. Table 2 shows the summary of water quality parameters retrieved using remote sensing, associated sensors, retrieval approaches, limitations, and data requirements.

3.3.1 Optically Active Parameters

The determination of Chl-a through satellite RS has been an active area of research due to its ecological negative impact when there are high concentrations during bloom events [16]. Latest research shows significant development in satellite-based Chl-a(Chl-a) retrieval across diverse aquatic systems through physics-based, semi-empirical, and machine-learning approaches. For example, Di Francesco et al. [8] applied semi-empirical spectral indices to improve Chl-a interpolation in small lakes, while Kowe et al. [108] revealed spatial eutrophication patterns via Sentinel-2. Similarly, Salem et al. [72] showed an enhanced Chl-a retrieval performance across low to high concentration conditions. Moreover, hybrid modeling strategies have shown favorable adaptability: Kandasamy et al. [109] integrated deep learning with the QAA algorithm to improve Chl-a prediction across river systems, whereas Tavares et al. [110] found that WaterDetect combined with improved AC produced consistent Chl-a estimates in small lakes. Physics-based QAA/GSM algorithms remained successful in optically complex waters, as demonstrated by Choto et al. [70] in eutrophic systems. Ensemble learning approaches are also gaining attraction; for example, Assaf et al. [111] found that Random Forest outperformed previous machine-learning models for Chl-a prediction in dams by detecting nonlinear spectral interactions. DNN also displayed outstanding results, with Najafzadeh and Basirian [51] reporting high retrieval accuracy for both Chl-a and SDD. Linear models remain useful for seasonal and thermal dynamics, as Dong et al. [98] demonstrated, capturing seasonal Chl-a variability using regression models. In coastal systems, Chen et al. [112] successfully used support vector regression to retrieve Chl-a along with dissolved oxygen and chemical oxygen demand. Jointly, these studies show the growing potential of machine-learning, semi-analytical/absorption-based, and atmospheric-correction-enhanced models for Chl-a retrieval, supporting improved monitoring of trophic status, spatial heterogeneity, and temporal variability across inland and coastal waters.

Several recent studies applied RS and modeling approaches for estimating SDD in inland and coastal waters. Rivera-Ruiz et al. [65] reported that the C2XC algorithm produced the most accurate SDD predictions in eutrophic lakes, specifically under high trophic conditions. Schmidt et al. [64] combined SDD and Chl-a retrievals and revealed that it can support water-quality assessments. Joshi et al. [94] found that RF models successfully predicted SDD for bloom-monitoring applications, while Najafzadeh and Basirian [51] showed that DNN improved SDD prediction accuracy using Sentinel-2 reflectance inputs. In optically complex waters, Qing et al. [66] indicated that improvements to the QAA algorithm enhanced SDD inversion performance. To sum up, these results indicate that both physics-based and machine-learning algorithms continue to advance the retrieval of SDD across diverse water conditions.

One of the most crucial factors in evaluating the quality of water is turbidity, which is also a reliable predictor of the overall number of suspended sediments. RS technologies have the potential to offer an affordable substitute for experimental methods that need a lot of time for field data collection and laboratory analysis in order to evaluate water turbidity over vast areas [37]. The improved prediction performance was observed across diverse water conditions as advances in turbidity retrieval have increased using the combination of satellite-derived reflectance and ML methods. For example, supporting reservoir monitoring applications, Assegide et al. [50] revealed reliable turbidity estimation from Sentinel-2 regression models, while Eljaiek-Urzola et al. [53] developed an empirical model for turbidity estimation using Sentinel-2 images and confirmed a robust retrieval performance using simple band-based reflectance. Ahmad et al. [49] reported that polynomial regression explained 87% of the variance in turbidity derived from Landsat reflectance, indicating strong predictive potential of nonlinear spectral relationships. Steinbach et al. [113] employed RF and GB, showing that turbidity predictions were strongly affected by climate–land cover interactions. In coastal environments, Kong et al. [114] found that RF achieved the highest turbidity retrieval accuracy among the tested ML models, highlighting its suitability for complex marine conditions. The continued integration of empirical regression and machine learning–based retrieval models, particularly those leveraging Sentinel and Landsat imagery, shows a continuing advancement in turbidity monitoring across a wide range of aquatic systems.

Recent advances in total suspended matter (TSM) and suspended particulate matter (SPM) retrieval illustrate a growing shift toward machine-learning and data fusion approaches. Zhang et al. [83] demonstrated that the ESTARFM–PLS framework improved spatiotemporal resolution of SPM predictions by effectively combining multi-source satellite imagery. Using Himawari-8 observations, Patricio-Valerio et al. [115] showed that artificial neural networks generated accurate TSS retrievals, highlighting the capability of geostationary platforms to support high-frequency monitoring. Similarly, Maniyar et al. [116] reported that DNN best predicted TSS concentrations (R2 = 0.89), outperforming RF and XGBoost, suggesting that deeper architectures may better capture nonlinear reflectance–TSS relationships. In nearshore environments, Igoe et al. [85] demonstrated that XGBoost could estimate suspended sediment concentrations while resolving spatial–temporal variability, showing the ability for dynamic coastal systems. Further, Chen et al. [117] found that DFE-ML improved TSM retrieval by learning nonlinear spectral representations, reinforcing the value of advanced ML feature extraction for sediment-laden waters. These results indicate that deep learning, ensemble ML, and data fusion strategies are advancing SPM/TSS retrieval accuracy and spatial–temporal continuity for coastal and inland water quality monitoring.

The reviewed studies also expanded the application of RS for monitoring thermal-related water quality indicators and their ecological effects. For instance, Ahmed et al. [48] reported that Landsat-derived surface temperature supported low-cost thermal monitoring and facilitated WQI calculation, determined the value of freely available multispectral sensors. Also, Adilakshmi & Venkatesan [47] showed that LASSO regression provided high-accuracy retrieval of water temperature, suggesting that feature selection enhances thermal signal extraction from spectral bands. ML approaches have also advanced thermal anomaly detection, as Luo et al. [82] showed that ML-based models supported real-time lake surface water temperature (LSWT) monitoring, critical for rapid response to heat-stress events. Further highlighting climate–water quality linkages, Kim et al. [75] observed dissolved oxygen decline associated with sea-surface temperature warming retrieved from MODIS/VIIRS imagery. Overall, these studies emphasize that the combination of multispectral satellite data, advanced regression, and ML models provides reliable tools for retrieving water temperature and detecting ecosystem responses tied to thermal variability.

Among other WQPs, CDOM was also observed, for example, Pelevin et al. [89] retrieved CDOM using LIDAR + Sentinel-2 + OC models for lake water, Zhang et al. [118] captured CDOM optical variability and source using MODIS models, and Choto et al. [70] retrieved CDOM with strong accuracy in eutrophic lake waters using absorption-based IOP algorithms.

Salinity retrieval using RS and machine-learning approaches has also been examined. Bygate and Ahmed [87] reported that salinity could be successfully retrieved using ML techniques, with deep neural networks achieving the highest performance following calibration. This result suggests that nonlinear learning models may better capture the indirect spectral relationships associated with salinity variability, particularly when supported by appropriate calibration datasets.

3.3.2 Optically Inactive Parameters

Recent developments in dissolved oxygen (DO) retrieval demonstrate the growing capability of satellite data and machine learning to support aquatic ecosystem assessments. Cao et al. [46] showed that Sentinel-2 regression models enabled accurate DO concentration inversion, highlighting the utility of freely available multispectral imagery. Similarly, modeled DO from Landsat reflectance to support Water Quality Index estimation, emphasizing RS as a cost-effective alternative to in situ sensing networks. Dong et al. [119] reported that ML-based DO retrieval captured significant interannual variability, demonstrating sensitivity to long-term environmental signals. In marine waters, Chen et al. [112] demonstrated that support vector regression successfully modeled nonlinear DO responses, reinforcing the advantage of non-parametric models in dynamic oceanic settings. Additionally, Kim et al. [75] monitored DO decline using sea surface temperature–linked satellite proxies, underscoring the importance of RS for detecting climate-induced deoxygenation trends. These advancements suggest that DO retrieval is no longer limited by sensor constraints and by synthesizing the satellite reflectance with thermal proxies, ML-based retrievals provide increasingly reliable DO estimation essential for regional to global water-quality monitoring.

Moreover, recent work has explored the probability of estimating biochemical oxygen demand (BOD/BOD5) from satellite data and machine-learning techniques. Dawn et al. [79] reported that ML-based models retrieved BOD with a very high accuracy of R2 = 0.99, which indicates an effective predictive ability. However, it should be interpreted with caution. In this result, the model with the highest accuracy includes all the spectral and water quality indices, land cover data, and meteorological data (precipitation, temperature, and wind speed). Compared to the scenarios using RS indices only and RS indices plus meteorological data, these two scenarios show lower accuracy results. It can be interpreted that sample sizes strongly influenced the accuracy of the report. An independent validation across both space and time is also required for a robust spatiotemporal evaluation, for example, on the variation in the number of in situ observations that may influence the model’s ability to generalize across different water bodies.

Similarly, Fu et al. [120] found that BOD5 retrieval performance exceeded other optically active parameters when using ML approaches, suggesting that nonlinear models can capture indirect spectral relationships to biochemical demand. Also, Arias-Rodriquez et al. [121] presented that ML relatively predicted BOD in coastal surface waters, pointing to potential applications for large-area monitoring while also reflecting remaining uncertainty in optically complex settings. While the ML-based retrieval may provide a transferable framework for BOD estimation, the transferability of these models remains contingent on the environment, such as variation with optical complexity and water type.

Chemical Oxygen Demand (COD) retrieval using RS and machine-learning approaches has received increasing attention. For example, Dong et al. [119] reported that CDOM-linked spectral modeling enabled COD retrieval in remote catchments, illustrating the potential of indirect optical proxies when direct absorption features are weak. This implies that the success of COD monitoring often depends on the model’s ability to accurately present the body’s organic composition. The trend toward using these indirect proxies has been further strengthened by using multi-parameter frameworks. For instance, Cao et al. [46] showed that CODMn could be estimated from Sentinel-2 reflectance with comparatively strong predictive performance. However, the complexity of aquatic conditions often requires more than a single-variable focus. To address this complexity, in a marine context, Chen et al. [112] demonstrated that support vector regression retrieved COD jointly with dissolved oxygen and phosphate, indicating that multi-parameter estimation frameworks may help capture linked biogeochemical interactions. While COD retrieval is feasible using multispectral reflective and ML approaches, the primary constraints remain on the performance stability of proxy selection and water optical conditions.

RS-based total TP retrieval from recent studies highlights the progressively improved capability of ensemble and hybrid machine-learning models. For instance, Sun et al. [58] demonstrated that bagging ensemble approaches reduced TP inversion bias and improved prediction accuracy, highlighting the aid of variance reduction through bootstrap aggregation. Similarly, Zhang et al. [67] applied XGBoost to predict TP dynamics and successfully detected long-term decline trends, illustrating the utility of boosted learners for capturing subtle nutrient trajectories. Taking advantage of Sentinel-2 reflectance, Dong et al. [119] confirmed that machine-learning models efficiently characterized strong seasonal and annual TP variations. Hybrid model integration also proved to be beneficial; for example, Hu et al. [122] reported that GA-ML fusion enhanced TP inversion across multisensor imagery by optimizing feature selection and model parameters, Zhou et al. [57] demonstrated that optimized XGBoost achieved the highest TP retrieval accuracy among tested models, reinforcing its capability for nutrient concentration prediction. Overall, these studies signify that ensemble learning, genetic algorithm–based optimization, and multisensory data fusion are driving developments in TP retrieval reliability, temporal trend detection, and spatial transferability for inland water quality monitoring.

The RS–based retrieval of nitrogen nutrients has evolved in aquatic systems using multispectral imagery and machine-learning approaches. For example, Rodríguez-López et al. [55] retrieved TN in lakes using spectral index regression, showing that reflectance-derived proxies can capture nutrient variability. Another work by Keith [123] demonstrated the feasibility of TN retrieval using Landsat-3 MSS, jointly estimating TN and Chl-a from multispectral reflectance. However, modern approaches have significantly enhanced this accuracy, such as the work of Hu et al. [122], which advanced the TN modeling by integrating genetic algorithms with RF, achieving high prediction accuracy from multi-sensor imagery, including GF-1, Landsat-8, and Sentinel-2, highlighting the benefits of feature optimization and cross-platform integration. This methodological shift establishes that feature optimization is now essential for capturing nutrient variability across diverse aquatic systems.

More recently, Arias-Rodriguez et al. [121] reported moderate prediction performance of machine-learning models for nitrate-N (NO3-N), albeit with lower accuracy compared to SDD, turbidity, and BOD retrievals, pointing to the challenge of detecting nutrients with weak optical signatures. To overcome this local performance gap, recent efforts have shifted toward global applications, Zhong et al. [78] developed a global nitrate estimation model validated across multi-sensor sea-surface reflectance datasets, demonstrating improved model transferability across regions and oceanographic conditions. The implication is that it indicates that nitrogen retrieval remains challenging but feasible when leveraging spectral indices, optimized ML, and multisensory fusion, with emerging efforts enhancing spatial scalability and prediction reliability for monitoring nutrient enrichment.

Across both optically active and proxy-inferred parameters, retrieval performance is strongly conditioned by AC quality, adjacency effects from surrounding land, and shallow-water bottom reflectance, particularly in narrow rivers, reservoirs, and coastal zones. These factors introduce systematic bias and reduce model transferability if not explicitly addressed. While machine-learning approaches can partially compensate for spectral limitations, their generalization remains constrained by training data representativeness and validation design. Consequently, robust RS-based water quality monitoring requires careful parameter classification, uncertainty reporting, and independent spatial–temporal validation, especially when inferring optically inactive constituents.

3.3.3 Key Gaps and Priorities

Key gaps

•   Over-reliance on empirical correlations for optically inactive parameters (e.g., DO, TN/TP, BOD/COD), with limited mechanistic justification and constrained transferability across regions and seasons.

•   Insufficient treatment of confounding factors, including atmospheric correction uncertainty, adjacency effects, shallow-water bottom reflectance, and proxy co-linearity among Chl-a, turbidity, and temperature.

•   Inadequate validation design, with frequent use of random cross-validation and small sample sizes that risk over-optimistic performance estimates.

•   Limited reporting and propagation of uncertainty, particularly when RS-derived parameters are aggregated, interpolated, or combined into indices.

•   Scale mismatch between RS products and application needs, especially in narrow rivers, small reservoirs, and fragmented inland waters.

Research priorities

•   Explicit separation of robust retrieval and proxy-based estimation, with clear labeling of empirical models as context-dependent unless independently validated.

•   Improved validation strategies, emphasizing spatial–temporal independence, multi-site testing, and uncertainty quantification.

•   Integration of physically informed and hybrid physics–ML approaches to improve robustness and interpretability of retrievals.

•   Multi-sensor and multi-resolution data fusion, combining satellites, UAVs, and in-situ measurements to address scale and coverage limitations.

•   Uncertainty-aware RS–GIS workflows, ensuring that uncertainty and resolution constraints are transparently communicated in decision-support applications.

Evidence of reporting bias due to missing results was observed across the reviewed studies, particularly in cases where model performance was reported without enough detail on validation strategies, dataset limitations, or unsuccessful model outcomes. Many studies emphasized high prediction accuracy but offered limited information on model uncertainty, negative results, or comparison with alternative methods. This reporting may introduce systematic overestimation of model performance and limit the robustness of synthesized conclusions.

3.4 Retrieval Methods

Across all the reviewed studies, retrieval methods span a continuum from simple empirical band ratios to advanced deep-learning, data fusion, and time-series frameworks. Table 3 shows the summary of the retrieval approaches and observation platforms used in this review.

3.4.1 Empirical Band Ratios, Spectral Indices, and Classical Regression

The first large group relies on empirical relationships between reflectance and water quality. These empirical approaches are easy to implement and often perform strongly in site-specific calibrations, but they tend to be region- and sensor-dependent and may struggle with transferability to other water types. These methods are more physically interpretable and often more transferable across sites than purely empirical regressions, but they require careful parameterization of IOPs and often need local tuning. Many studies develop uni- or multivariate regression models from individual bands or band combinations to retrieve turbidity, nutrients, and other WQPs (e.g., one-dimensional regression on NIR bands for DO, TN, TP, and turbidity in Sentinel-2 data) by Cao et al. [46]; OLS and polynomial regression for turbidity from Landsat-8 band combinations by Ahmad et al. [49]; multiple linear regression for DO, Chl-a, TSM, and thermal effluents by Dong et al. [98]; and multiple linear or multivariate regression for nitrate and water color by Zhong et al. [78]. Several works build empirical algorithms using specific band ratios or indices: B5/B4, B4/B3, and single-band models for Chl-a, turbidity, and TSS from Sentinel-2 [50]; NDVI-based empirical models for TN, turbidity, Chl-a, and TSS [108,124]; red-band empirical turbidity and SDD models from Sentinel-2 [53]; multi-band indices such as NDCI, NDAI, MCI, and CI from drone RGB/multispectral imagery [100]. LASSO regression is used to select informative bands/indices and build parsimonious predictive models for multiple parameters (temperature, TDS, pH, turbidity, Chl-a, DO, BGA) from Landsat-8 in rivers [47,48]. Multivariate Adaptive Regression Splines (MARS), Gene Expression Programming (GEP), and Evolutionary Polynomial Regression (EPR) are applied to derive WQI or individual WQPs from Landsat-8 [51].

3.4.2 Semi-Analytical and Analytical Bio-Optical Algorithms

A second family uses physics-informed, semi-analytical, or fully analytical models linking inherent/optical properties to constituent concentrations: (a) The classic QAA (e.g., QAA v5/v6) is used to retrieve Chl-a and other IOPs from satellite reflectance, sometimes coupled to AI: for Chl-a from Landsat/MERIS [109]; for Secchi depth (ZSD) in lakes [66]; and in a suite of IOP algorithms (QAA_v6, GSM, GIOP-DC, PML, LMI) to derive specific absorption terms and then water quality variables such as Chl-a and suspended solids [70]. (b) Ocean-color style algorithms are widely used: OC2 and Nechad algorithms for Chl-a and TSS from Sentinel-2 and in situ LIDAR/S2 data [89,92]; the JAXA Chl-a Version-2 algorithm, which merges a color index algorithm for low Chl-a with OCx algorithms for high Chl-a and also derives TSM [72]. (c) ESTARFM-based spatiotemporal fusion is combined with PLS regression to retrieve SPM at improved spatial/temporal resolution [83]. Semi-empirical band-index models are used with ACOLITE-processed PlanetScope and Sentinel-2 to estimate Chl-a, turbidity, and cyanobacteria in lakes [8], or to retrieve Chl-a and DOC/CDOM in lakes [118]. Semi-analytical Chl-a retrieval from multi-sensor ocean color data is used for coastal hazard monitoring [77] and for Sentinel-3 OLCI in lakes [73].

3.4.3 Machine Learning Regression

Additionally, to date, the most widely used RS estimation model is based on cutting-edge ML/AI technology. A large group of studies use supervised machine learning to map spectral features (bands, indices, or IOPs) to WQPs. These methods exploit non-linear relationships in feature space, often outperforming traditional empirical or semi-analytical algorithms, but they are highly sensitive to training data quality and representativeness. RF and GB are used for turbidity prediction from Sentinel-2 plus ancillary data [113]; turbidity and nutrients from Sentinel-2/Landsat [40,82,100]; Chl-a, TN, TP, and multi-parameter lake eutrophication indices [94,95,125,126]; CDOM and DOC [127]; TSM and TSS in coastal/marine waters [60,97]. Gradient boosting trees or boosted regression trees are also used to link long-term Chl-a/turbidity trends to catchment pressures [90]. XGBoost is a dominant retrieval method for Chl-a, TP, TN, turbidity, and LSWT: e.g., turbidity retrieval with MSI/OLI [57]; TN and TP using MODIS-derived reflectance [67]; algal biomass abundance [128]; hybrid GA-XGB models for TP from UAV imagery [43]; TP inversion from Sentinel-2 [56,57]; LSWT forecasting from proximal hyperspectral data [82]; and coastal Chl-a mapping from combined Sentinel-1/2/3 [129]. CatBoost appears both as a stand-alone regressor and within hybrid or ensemble frameworks [109,111,119]. SVR is used for Chl-a and turbidity retrieval from OLCI, HLS, and combined satellite products [127,130]; SVM is applied to retrieve pH, DO, TSS, and TDS from Sentinel-2/ResourceSat-2 [131] and to support algal bloom prediction [96].

3.4.4 Neural Networks and Deep-Learning Frameworks

Beyond classical ML, many studies deploy ANN and deep networks. These deep-learning approaches provide high accuracy and explicit uncertainty estimates (in the Bayesian/MDN cases), but often require large, diverse training datasets, which the studies repeatedly identify as a key limitation. Standard ANN and MLP-based DNNs are used for TDS and TSS from Landsat-8 [97,132], for Chl-a/TSM and CyanoHABs from MERIS/OLCI [71,133], and for multi-parameter lake water quality [52]. CNNs and deep feature extraction–ML fusion frameworks are used to extract spatial–spectral features and then feed SVR/RF regressors for TSM [117], to delineate algal bloom area [134], and to estimate Chl-a from Sentinel-2 MSI in Poyang Lake [135]. Deep Feature Extraction + ML (DFE-ML) and transfer-learning-based networks (e.g., TLNet) further appear in UAV-based urban river monitoring [101]. Time-aware LSTM and RNN architectures are implemented to model temporal Chl-a dynamics from Landsat-8 [55] and for forecasting eutrophication states or LSWT when combined with meteorological forcing. NBeats time-series models are coupled to CatBoost and QAA in [109] for time-series chlorophyll retrieval. Bayesian probabilistic neural networks (BNNs) are used to retrieve Chl-a and quantify uncertainty across OLCI/MSI [136], while Mixture Density Networks (MDNs) are used to retrieve cyanoHAB magnitude and probability from Sentinel-2/3 data [84].

3.4.5 Data Fusion, Data Assimilation, and Optimal Estimation

Several studies treat retrieval as a data assimilation or fusion problem. These methods integrate multi-sensor EO data with process-based models, enabling gap-filling and higher temporal resolution, but they require significant modelling infrastructure and expert knowledge. MERIS Chl-a products are assimilated into a coastal model using Ensemble Kalman filtering and smoothing to generate daily Chl-a fields in the Baltic Sea (MERIS–EnKF). CLMS and ESA CCI lake temperature, turbidity, and Chl-a are retrieved using Optimal Estimation from AATSR, SLSTR, MERIS, and OLCI L1 data [79]. Enhanced spatiotemporal reflectance products are created via ESTARFM [83] or via particle-tracking/numerical hydrodynamic models [137], after which regression/PLS is used to retrieve SPM and TSM.

3.4.6 Retrieval from Unmanned Platforms (UAVs, USVs, Proximal Sensors)

Finally, a group of studies uses proximal sensing platforms with retrieval methods like satellite-based ones. These studies extend retrieval techniques to very high spatial/temporal resolution, but note limitations in spatial coverage, cost, and seasonal representativeness. GA-optimized XGBoost (GA-XGB) for TP from UAV multispectral imagery [43]; TLNet, CNN, LSTM, Transformer, and conventional ML for TN, DO, TSS, and Chl-a in urban rivers [101]; random forest, GBM, XGB, and KNN for E. coli from drone RGB imagery [99]. Hyperspectral proximal sensing (HPSs) combined with XGBoost, DNN, and KNN is used to forecast LSWT at hourly scales [82]. USV-mounted sensors produce dense point measurements that are interpolated via IDW/Kriging [44].

3.4.7 Geostatistical Interpolation and GIS-Integrated Regression

A smaller group employs spatial interpolation or GIS–statistical coupling. Although not “spectral inversion” in the strict sense, these interpolation-type methods are retrieval procedures for spatial fields of water-quality indicators. Kriging and inverse distance weighting (IDW) are used to interpolate Chl-a from in situ measurements collected by unmanned surface vehicles (USVs), generating spatially continuous Chl-a maps [44]. Stepwise regression combined with RS-derived indices in a GIS environment is used to build predictive models for multiple physico-chemical parameters in lakes [138]. Across the reviewed studies, retrieval approaches have evolved along three main trajectories: (1) from simple empirical regressions to sophisticated ML and deep learning. Early and simpler studies rely on band ratios, indices, and linear/polynomial regression, which remain valuable for transparent, low-data settings. However, most recent work favours tree-based ensembles (RF, XGBoost, CatBoost), ANNs/DNNs, CNNs, and LSTMs because they better capture non-linear and multivariate relationships, especially for complex or non-optically active parameters (TN, TP, DO, BOD, nutrients, E. coli). These methods routinely outperform empirical and even some semi-analytical algorithms, but depend heavily on large, well-distributed training sets and careful cross-validation, (2) from purely data-driven to hybrid physics–ML and fusion frameworks. Semi-analytical/bio-optical algorithms (QAA, GSM, Nechad, OCx/CI) and sophisticated AC (ACOLITE, C2RCC, GRS) remain central, particularly where transferability and physical interpretability are priorities. A clear trend is to combine these physically based products with ML (e.g., QAA-derived IOPs used as ML features, ESTARFM + PLS, EnKF, or OE-based data assimilation feeding ML predictors), enabling robust, high-frequency retrievals while preserving physical consistency, and (3) towards spatiotemporal, multi-platform, and uncertainty-aware retrieval. Increasingly, studies treat retrieval as a spatiotemporal forecasting and data-fusion problem: time-series models (ARIMA, NBeats), BRT trend analysis, LSWT and Chl-a forecasting, and multi-sensor fusion across MODIS, OLCI, MSI, OLI, UAVs, USVs, and proximal sensors. Probabilistic deep networks (BNN, MDN) explicitly estimate uncertainty, and data assimilation frameworks address observational gaps. Unmanned platforms and proximal hyperspectral systems push towards real-time, high-frequency monitoring, but their models still face challenges in scaling beyond individual sites or seasons.

Fig. 2 illustrates a hierarchical framework of retrieval methods underpinning RS–based water quality monitoring. The framework demonstrates the progressive evolution of retrieval approaches—from simple empirical regressions to semi-analytical ocean-color algorithms, machine learning and deep learning models, data assimilation/time-series fusion strategies, and emerging unmanned and proximal sensing systems. This hierarchy emphasizes increasing capability, scalability, and uncertainty awareness in retrieving key water quality parameters. By mapping these methodological tiers to their resulting applications in inland and coastal monitoring, early-warning systems, and policy decision support, the figure highlights how advances in retrieval science contribute to more accurate, continuous, and operational water quality assessment.

Figure 2: Hierarchical framework of RS retrieval methods and their applications in surface water quality monitoring.

3.5 GIS Integration in Water Quality Monitoring

A Geographic Information System (GIS) is an important aspect of environmental monitoring because it combines geospatial data and environmental data to conduct spatial analyses as well as gain a visual aspect [139]. GIS can help identify water quality, locate discharge sites for industrial wastewater, examine the dispersion of pollutants, and help designate areas of concern [140]. GIS can assist in facilitating pollution risk mapping, assist in hydrological modeling, and support assessments of pollution in transportation dimensions in river ecosystems [141].

3.5.1 Conceptual RS–GIS Integration Framework

Geographic information systems provide techniques for storing, processing, and controlling a large amount of information derived from Remote Sensing Devices [138]. In the reviewed literature, GIS functions as a critical integrative layer that links remotely sensed (RS) water quality products with spatial context, ancillary datasets, and decision-support workflows. A generalized RS–GIS integration framework can be conceptualized as a three-stage process: (i) RS-derived parameter retrieval, where water quality parameters (e.g., chlorophyll-a, turbidity, total suspended solids, nutrients) are retrieved from satellite or UAV imagery using empirical, semi-analytical, or machine-learning approaches [15,106]; (ii) GIS-based spatial analysis and inference, and (iii) management-oriented interpretation. These spatially continuous RS products are subsequently imported into GIS environments, where they are combined with hydrological networks, land-use data, administrative boundaries, and in-situ observations to enable spatial inference and visualization [16].

This framework highlights that GIS does not independently generate water quality information, but rather facilitates interpretation, scaling, and decision relevance of RS-derived estimates. Consequently, the reliability of RS–GIS integration depends strongly on how spatial resolution, uncertainty, and hydrological structure are treated across these stages.

3.5.2 Typical GIS-Enabled Inference Modes

GIS provides several recurring inference modes within the reviewed studies. The most widely used approach is spatial interpolation and surface modeling, where point-based in-situ measurements or RS-derived retrievals are interpolated using methods such as inverse distance weighting (IDW) or kriging to infer continuous water quality maps. While effective for visualization, these approaches assume spatial stationarity relationships, which may not reflect hydrologically driven transport processes.

A second inference mode involves zonal statistics and overlays analysis, where RS-derived water quality parameters are aggregated within hydrological units, administrative boundaries, irrigation command areas, or reservoir extents [8,48]. This approach is commonly applied in water quality compliance monitoring and reporting but may mask sub-pixel variability and localized pollution gradients when medium-resolution RS products are aggregated to fine management units.

Another mode is index-based synthesis, particularly through GIS-implemented Water Quality Index (WQI) mapping. WQI integrates multiple parameters into a single score to support communication and decision-making [26]. However, several studies note that GIS-based WQI maps can obscure parameter-specific exceedances, amplify uncertainty when RS-derived proxies are used, and create an illusion of spatial precision that exceeds the robustness of the underlying inputs [27].

More recent work combines machine-learning–assisted GIS inference, where RS products, terrain attributes, land use, and hydrological variables are jointly analyzed within GIS environments to identify pollution drivers, hotspots, or vulnerability zones [54]. Despite its potential, these approaches are sensitive to training data representativeness and spatial autocorrelation.

3.5.3 Scale, Uncertainty, and Connectivity Considerations

A central limitation identified across the literature is scale mismatch between RS products and GIS layers. Medium-resolution satellite data (10–30 m) are frequently integrated with fine-scale GIS features such as narrow river networks, point-source locations, or irrigation canals, leading to spatial averaging effects and misrepresentation of localized water quality gradients [142]. This issue is particularly pronounced in narrow rivers, small reservoirs, and fragmented inland waters, where mixed land–water pixels and adjacency effects further degrade retrieval accuracy [103,143].

Uncertainty propagation is another persistent challenge. Many GIS-based studies visualize RS-derived products without explicitly communicating retrieval uncertainty, validation limits, or sensitivity to atmospheric correction and proxy co-linearity. When these products are subsequently interpolated, aggregated, or combined into indices such as WQI, uncertainty can compound and become difficult to trace.

Finally, hydrological connectivity is often inadequately represented in GIS analyses. Interpolation and overlay methods commonly neglect flow direction, upstream–downstream relationships, residence time, and mixing processes that govern water quality dynamics in river networks and irrigation systems [16]. As a result, spatial proximity in GIS does not necessarily correspond to hydrological influence, limiting the interpretability of mapped patterns.

3.5.4 Implications for Operational RS–GIS Integration

Moving beyond simple data layering, the reviewed evidence indicates that GIS integration in water quality monitoring should be treated as a decision-support and interpretation layer, rather than a purely analytical extension of RS retrieval. Robust RS–GIS integration requires explicit alignment between sensor resolution and GIS analysis scale, careful selection of inference modes, and transparent reporting of uncertainty and validation constraints. Emerging approaches that combine multi-resolution RS data, hydrologically informed GIS analysis, and uncertainty-aware visualization show promise but remain underdeveloped in operational practice.

Table 4 shows the summary of the synthesized and reviewed literature according to study objectives, which highlights how recent studies contribute to addressing limitations of traditional monitoring and advancing scalable RS–GIS-based water quality assessment frameworks.

4  Discussion

4.1 Interpretation of Results in Relation to Existing Literature

The synthesis indicates that RS-based surface water quality monitoring has progressed from a supplementary role to become a central spatial framework for environmental evaluation. Rather than a replacement for field sampling, RS addresses the limitation of localized observations by enabling spatially continuous and repeatable monitoring. This reflects a broader methodological shift in literature from point-based measurement toward system-scale environmental assessment.

The reviewed evidence demonstrated that a critical distinction emerged from concerns about optically active and optically inactive parameters. The foundational strength of the retrieval of optically active parameters is comparatively robust and reinforced by semi-analytical and empirical models with demonstrated regional applicability [106], while estimation of optically inactive parameters increasingly relies on machine-learning (ML) models that exploit indirect spectral proxies and nonlinear relationships. Although ML models often outperform traditional approaches in estimating optically inactive parameters, this performance reflects improved proxy modeling rather than true optical retrieval. As such, ML expands the inferential capacity of RS but does not eliminate the structural constraints associated with calibration dependency, transferability, and validation.

The transition from the reliability of single sensors moving towards the multi-sensor and multi-platform integration using satellite, UAV, thermal, radar, and in-situ observations also shows as a consistent pattern in RS water quality monitoring. This integration mitigates resolution trade-offs, reduces cloud-related gaps, and enhances temporal continuity [97,144]. This shows that RS-based monitoring is turning into an integrated observing system rather than a single-algorithm solution. By taking advantage of its complementary strengths across platforms, it may solve the data limitations and improve observational continuity.

The role of GIS further highlights this transition. When RS-derived parameters are integrated within geospatial workflows, such as hotspot mapping, watershed-scale interpolation, and scenario analysis, they shift from descriptive indicators to decision-support tools. This integration connects spectral retrieval, enabling spatially explicit environmental management. However, there are still determined constraints, such as atmospheric correction uncertainty, adjacency effects, shallow-water interference, and limited spatial–temporal validation, that continue to restrict model generalizability [102]. These factors may introduce systematic preference and increase sensitivity to local optical conditions, which reduces the robustness of retrieval algorithms across different water bodies.

Generally, the reviewed literature suggests that the field is transitioning from algorithm optimization toward system-level robustness. Future progress depends not only on improving retrieval accuracy, but on developing transferable, uncertainty-explicit, and scalable RS–GIS monitoring architectures.

4.2 Theoretical and Practical Implications

The findings from this review strengthen the theoretical framework of surface water quality monitoring by explicitly distinguishing the conceptual boundary between optically active and optically inactive water quality parameters. It is evident that while optically active parameters are governed by direct bio-optical interactions, optically inactive parameters are fundamentally proxy-based and dependent on indirect light-based measurement. The distinctions between these parameters have important theoretical implications as it highlights the retrieval uncertainty and transferability, which means that RS-based water quality monitoring is parameter-dependent rather than sensor- or algorithm-dependent. The results further contribute to the growing RS-based theoretical framework by emphasizing the role of ML as an extension to water quality monitoring, which supports the emerging hybrid paradigm of combining physical features and data-driven approaches to enhance robustness, generalization, and interpretability in water quality retrievals. In terms of sensors and platforms, the findings from the reviewed literature indicate that no single sensor can resolve all spatial, temporal, and spectral requirements of water quality monitoring. These results synthesize the evidence supporting that integrated observation systems, by combining satellite RS, UAVs, in-situ sensors, and GIS, represent a fundamental evolution toward scalable and uncertainty-aware monitoring frameworks.

From a practical perspective, the findings support that RS can extensively facilitate a paradigm shift from assessments from discrete point-based sampling to basin- and watershed-scale assessments. This capability is especially important for data-scarce regions with a lack of monitoring infrastructure, in which RS can provide cost-effective, repeatable, and spatially comprehensive information. As ML approaches demonstrates its effectiveness, they also have practical implications when compared to the traditional water quality monitoring, which is known to be time-intensive and expensive. This review highlights that using ML-based water quality monitoring requires careful calibration, validation, and uncertainty reporting. Lastly, the RS-GIS integration provides a robust framework for environmental governance and policy formulation through spatial visualization, hotspot identification, trend analysis, and scenario evaluation. This workflow can substantially support evidence-based decision-making in watershed management, pollution control, and land-use planning.

4.3 Gaps, Inconsistencies, and Future Research Directions

This review highlights significant progress in the application of RS and GIS for surface water quality monitoring; however, a synthesis of the literature reveals several gaps, inconsistencies, and challenges that limit operational deployment and scientific generalization. This review is also subject to several methodological limitations. The literature search was limited to the selected databases and types of publications, which may not capture other relevant data from other sources. Additionally, a qualitative synthesis was performed due to the significant heterogeneity in study designs, sensors, and analytical approaches, which precluded quantitative meta-analysis.

Despite the advancements highlighted in this review, the available evidence presents several limitations that affect the overall confidence in the findings. Several studies rely on empirical and machine learning–based models that are calibrated using limited or site-specific datasets, which may limit their transferability across different regions and varying environmental conditions. The extensive use of indirect proxy relationships for optically inactive parameters further introduces uncertainty due to weak spectral signals.

Behind these limitations, future research directions have been recommended for a robust decision-support tool in water quality monitoring.

4.3.1 RS Retrieval Gaps and Inconsistencies

Despite significant advancements in RS applications for water quality monitoring, the reviewed studies show various technical and methodological challenges and limitations that affect retrieval accuracy, spatial–temporal representativeness, and operational transferability. Various sensors are limited by insufficient resolution to capture localized-scale spatial heterogeneity in rivers, small lakes, or fragmented water bodies. For example, the 30 m resolution of Landsat imagery was insufficient for narrow inlets and complex lake fringes in BOD and TSS retrieval [79], while low sample density in river networks constrained accurate pollution mapping [145]. Similarly, limited revisit frequency in Sentinel-2 imagery resulted in temporal gaps that prevented continuous intra-seasonal monitoring [137]. Numerous optical RS studies emphasize signal degradation caused by atmospheric aerosols, adjacency effects from surrounding land, water-surface sun glint, and cloud-induced pixel contamination. For instance, adjacency effects have been shown to alter near-infrared (NIR) reflectance, thereby reducing the accuracy of lake water quality retrievals [110], while inconsistent pixels from cloud and ship shadows affected turbidity and TN/TP uncertainty [56]. These studies highlight the ongoing need for improved AC algorithms tailored to inland and coastal waters.

An extensive constraint is the insufficient number, coverage, and spatiotemporal alignment of ground measurements for model training, calibration, and validation. Work on coastal turbidity by [138] noted limited in-situ matchups for validating C2RCC retrievals. Similarly, sparse high-concentration nutrient samples reduced high-end predictive accuracy in TP/TN studies using XGBoost [67]. These constraints hinder model generalizability and confidence in extreme-value pollution assessment. Many retrieval models perform well only for a single region, season, water type, or algorithm–sensor pairing. For example, semi-empirical procedures for PlanetScope could not be directly transferred to other reservoirs [8], and algorithm performance varied with trophic state levels for Secchi depth retrievals [65]. This exposes the need for region-independent and multiscale transferable retrieval frameworks.

Several parameters (e.g., nutrients, metals, COD, salinity) exhibit weak spectral signatures, leading to higher uncertainties relative to optically active variables such as Chl-a or turbidity. The COCTS SPM products suffered algorithm limitations compared to Chl-a estimates, necessitating algorithm refinement. Similarly, Landsat sensor band limitations restricted Chl-a retrieval accuracy due to the absence of red-edge bands [146].

4.3.2 GIS Integration Gaps and Inconsistencies

While GIS techniques enhance spatial analysis, visualization, and decision support, the reviewed studies identify several limitations related to data sparsity, temporal coverage, and workflow consistency.

Several reviewed studies integrate medium-resolution satellite–derived water quality products (e.g., Landsat-8 OLI, Sentinel-2 MSI) with fine-scale GIS layers such as river networks, point-source locations, administrative boundaries, and management units. While this integration enhances spatial visualization and decision support, a recurring limitation identified in the literature is scale mismatch between satellite pixel resolution and the spatial granularity of GIS features. Medium-resolution pixels often spatially average reflectance signals over heterogeneous surfaces, which can obscure localized pollution gradients and introduce uncertainty when mapped onto narrow rivers, small reservoirs, or fragmented water bodies [15,106]. Also, the reviewed studies demonstrate that narrow river channels and small inland waters are frequently under-resolved by medium-resolution imagery, leading to mixed land–water pixels and biased water quality estimates when integrated into GIS frameworks [8,147]. This issue becomes particularly pronounced when GIS layers representing point sources or hydrological features are interpreted at scales finer than the effective resolution of the RS product, resulting in apparent spatial precision that exceeds the underlying data support.

The literature further indicates that scale mismatch interacts with adjacency effects and pixel contamination, especially in riverine and nearshore environments, compounding uncertainty in RS–GIS outputs [103,143]. Studies relying on GIS-based interpolation or zonal statistics without accounting for sensor resolution or mixed-pixel effects risk propagating these uncertainties into downstream analyses, including watershed prioritization and regulatory assessment.

Recent work highlights that higher-resolution commercial satellites, UAVs, or multi-sensor data fusion can partially mitigate scale mismatch, but these approaches remain limited by cost, coverage, and operational constraints [8,80]. Consequently, the reviewed evidence suggests that GIS integration should be accompanied by explicit discussion of scale compatibility and uncertainty, rather than treating satellite-derived products as directly comparable to fine-scale spatial layers.

GIS-based inverse distance weighting (IDW), kriging, or trend interpolation propagates error in data-scarce river basins. Several GIS-based suitability and WQI mapping studies highlight uncertainty linked to short monitoring periods [148] and sparse sampling networks lacking hydrochemical diversity [149]. The inability to capture long-term hydro-climatic variability weakens inference regarding seasonal or climate-driven processes. Combined RS + GIS workflows frequently suffer from inconsistent radiometry and AC between sensors, temporal mismatch, insufficient metadata, or a lack of harmonization workflows. Inter-sensor mismatch between MERIS/Sentinel-3 MSI/OLCI datasets undermined continuous Chl-a records and uncertainty quantification [136].

Overall, this critical review indicates that scale mismatch remains a fundamental limitation in RS–GIS-based water quality studies. Addressing this challenge requires careful alignment between sensor resolution and GIS analysis scale, transparent reporting of spatial uncertainty, and cautious interpretation of fine-scale patterns derived from medium-resolution satellite data.

4.3.3 Future Research Directions

A structured roadmap (Fig. 3) that outlines strategic priorities for advancing RS–GIS integration in surface water quality monitoring was developed from the technical challenges and emerging innovations identified across the literature. The roadmap positions the current limitations and aligned future directions across four interconnected pillars: (1) improving sensor, retrieval, and preprocessing foundations; (2) integrating multisource RS products; (3) standardized RS–GIS workflows for real-time, scalable deployment; and (4) embedding scientific outputs within governance and decision-support systems.

Figure 3: Proposed roadmap for advancing RS–GIS integration in surface water quality monitoring systems.

Pillar 1: Sensors, Retrieval, and Preprocessing Foundations

The consistent challenges in atmospheric correction, uncertainty, limited spatiotemporal resolution, validation, and calibration reliability continue to limit the model transferability. Therefore, robust retrieval and standardized preprocessing are needed to provide reliable water quality monitoring. The future directions also highlight long-term, multi-scale observation approaches. Works on the transferability of hybrid ML retrieval frameworks should integrate physically meaningful inputs (e.g., inherent optical properties, water-type classifications, or semi-analytical outputs) into ML architectures. A hybrid ML that upholds high predictive accuracy while improving interpretability and stability across sites, especially for proxy-inferred parameters, will represent a substantial methodological advance. Higher-quality calibration and prioritizing spatially and temporally independent validation strategies, such as independent regional validation, and not relying on a random data split. To do this, hybrid ML approaches, physics-informed retrieval, and standardized frameworks should prioritize clearly determining uncertainties. By considering these approaches, methodological robustness at this level may avoid model reliability limitations.

Pillar 2: Multi-Source and Multi-Resolution Integration

Many works emphasize the need to extend monitoring beyond short-term case studies by expanding spatial and temporal coverage, for example, GIS- and MCDA-based lake assessments recommend multi-year datasets to capture seasonal variability and ecosystem change [149] and increasing the number of monitoring stations to improve calibration and validation [146]. For example, future modeling suggests the integration of advanced ML techniques and multi-sensor datasets to obtain more scalable eutrophication and turbidity retrieval systems [62].

With the integration of multi-source and resolution, several works highlight the necessity of integrating operational implementation of water quality retrievals within RS-GIS-enabled planning tools and management frameworks for rivers, lakes, and coastal areas. Tighter coupling of RS–GIS analytics with water management decision systems, such as linking the RS-GIS derived results with regulatory standard thresholds and management indicators. By doing this, policy makers and water managers will be guided efficiently for early warning, compliance assessment, and planning.

Pillar 3: Embedding RS-GIS Standardized and Transferable Workflows

Despite strong predictive performance in many studies, reproducibility and generalizability remain limited across diverse environmental conditions. Operational deployment requires standardized preprocessing pipelines, open validation protocols, uncertainty reporting standards, and cross-basin model testing. Standardization transforms individual methodological advances into transferable monitoring systems. This pillar addresses the gap in experimental research and implementation. The ultimate value of RS-derived parameters lies in their integration into a spatial decision-support framework, such as GIS-based hotspot mapping, watershed-scale interpolation, scenario analysis, and policy-aligned indicators that convert spectral retrieval into an evidence-based water quality assessment [148]. Integrating RS outputs within geospatial structures shifts monitoring from descriptive mapping to adaptive water quality monitoring and management.

Pillar 4: Decision-Support

Taken together, these four pillars redefine RS–GIS monitoring as an integrated architecture roadmap to move from foundational spectral reliability to system-level integration, aligning methodological advancement with operational applicability.

5  Conclusion

This review combines advances in the integration of RS and GIS technologies for surface water quality monitoring and demonstrates the paradigm-shifting potential of these tools to overcome the challenges and limitations of traditional approaches to water quality monitoring and assessment, highlighting a clear methodological transition from site-specific, point-based observation toward integrated, multi-scale monitoring systems. Across the body of literature from 2012–2025, RS demonstrated a strong capability for the direct retrieval of optically active water quality parameters, including Chl-a, turbidity, TSS, CDOM, SDD, and surface water temperature. The evidence demonstrates that RS has progressed beyond a supplementary observational tool and now forms the backbone of spatially continuous environmental assessment. At the same time, the expansion of ML approaches has extended retrieval capabilities to parameters lacking direct optical signatures, though with increased demands for validation consistency and uncertainty transparency. GIS has proven necessary for spatial interpolation, watershed and catchment analysis, multi-criteria decision support, and integration of satellite-derived indicators with land use, hydrological, and climatic drivers.

Finally, the integration of the enhanced RS-GIS results with management metrics, pollution source tracing, water quality parameters, and policy decision tools is important to shift from proof-of-concept research towards continued operational monitoring. When supported by standardized workflows, uncertainty reporting, transparent calibration, and spatio-temporal validation practices, integrated RS–GIS systems can provide spatially explicit evidence suitable for structured decision-support processes in water-resource management. Addressing these gaps will enable the RS–GIS community to contribute to a more consistent, spatially explicit water quality monitoring that can inform environmental management and policy evaluation, subject to appropriate validation and uncertainty reporting.

Acknowledgement: Not applicable.

Funding Statement: The authors received no specific funding for this study.

Author Contributions: The authors confirm contribution to the paper as follows: Conceptualization, Jhoreene Julian and Jerome Gacu; methodology, Jhoreene Julian and Jerome Gacu; formal analysis, Jhoreene Julian; data curation, Jhoreene Julian and Jerome Gacu; writing—original draft preparation, Jhoreene Julian; writing—review and editing, Jerome Gacu; visualization, Jhoreene Julian; All authors reviewed and approved the final version of the manuscript.

Availability of Data and Materials: Not applicable.

Ethics Approval: Not applicable.

Conflicts of Interest: The authors declare no conflicts of interest.

Supplementary Materials: The supplementary material is available online at https://www.techscience.com/doi/10.32604/rig.2026.078160/s1.

References

1. Wang Y, Wu L, Wei Z, Tan T, Yu Y, Wang Z, et al. Coordinating food security and environmental performance in Belt and Road agri-food systems. Resour Conserv Recycl. 2025;219(9):108304. doi:10.1016/j.resconrec.2025.108304. [Google Scholar] [CrossRef]

2. Stets EG, Sprague LA, Oelsner GP, Johnson HM, Murphy JC, Ryberg K, et al. Landscape drivers of dynamic change in water quality of U.S. rivers. Environ Sci Technol. 2020;54(7):4336–43. doi:10.1021/acs.est.9b05344. [Google Scholar] [PubMed] [CrossRef]

3. Tian Q, Luo W, Guo L. Water quality prediction in the Yellow River source area based on the DeepTCN-GRU model. J Water Process Eng. 2024;59(19):105052. doi:10.1016/j.jwpe.2024.105052. [Google Scholar] [CrossRef]

4. Nong X, Shao D, Zhong H, Liang J. Evaluation of water quality in the South-to-North water diversion project of China using the water quality index (WQI) method. Water Res. 2020;178(15):115781. doi:10.1016/j.watres.2020.115781. [Google Scholar] [PubMed] [CrossRef]

5. Chen H, Gao X, Yuan R. Advances in remote sensing and sensor technologies for water-quality monitoring: a review. Water. 2025;17(20):3000. doi:10.3390/w17203000. [Google Scholar] [CrossRef]

6. Okpara EC, Sehularo BE, Wojuola OB. On-line water quality inspection system: the role of the wireless sensory network. Environ Res Commun. 2022;4(10):102001. doi:10.1088/2515-7620/ac9aa5. [Google Scholar] [CrossRef]

7. Di Francesco S, Venturi S, Casadei S. An integrated water resource management approach for Lake Trasimeno, Italy. Hydrol Sci J. 2023;68(4):630–44. doi:10.1080/02626667.2023.2185150. [Google Scholar] [CrossRef]

8. Di Francesco S, Biondi F, Casentini B, Fazi S, Amalfitano S, D’Eugenio M, et al. Sentinel-2 and Planet-Scope as reliable tools for water quality monitoring of small reservoirs. Egypt J Remote Sens Space Sci. 2025;28(4):713–23. doi:10.1016/j.ejrs.2025.11.002. [Google Scholar] [CrossRef]

9. Lukhabi DK, Mensah PK, Asare NK, Pulumuka-Kamanga T, Ouma KO. Adapted water quality indices: limitations and potential for water quality monitoring in Africa. Water. 2023;15(9):1736. doi:10.3390/w15091736. [Google Scholar] [CrossRef]

10. Chen P, Wang B, Wu Y, Wang Q, Huang Z, Wang C. Urban river water quality monitoring based on self-optimizing machine learning method using multi-source remote sensing data. Ecol Indic. 2023;146(12):109750. doi:10.1016/j.ecolind.2022.109750. [Google Scholar] [CrossRef]

11. Sun Y, Wang D, Li L, Ning R, Yu S, Gao N. Application of remote sensing technology in water quality monitoring: from traditional approaches to artificial intelligence. Water Res. 2024;267(1):122546. doi:10.1016/j.watres.2024.122546. [Google Scholar] [PubMed] [CrossRef]

12. Ooko SO, Cheptegei L, Karume SM. Application of machine learning for real-time water quality monitoring in developing countries: a review. Sustain Futur. 2025;10(4):100984. doi:10.1016/j.sftr.2025.100984. [Google Scholar] [CrossRef]

13. Liang Y, Ding F, Liu L, Yin F, Hao M, Kang T, et al. Monitoring water quality parameters in urban rivers using multi-source data and machine learning approach. J Hydrol. 2025;648(23):132394. doi:10.1016/j.jhydrol.2024.132394. [Google Scholar] [CrossRef]

14. Nguyen UNT, Pham LTH, Dang TD. An automatic water detection approach using Landsat 8 OLI and google earth engine cloud computing to map lakes and reservoirs in New Zealand. Environ Monit Assess. 2019;191(4):235. doi:10.1007/s10661-019-7355-x. [Google Scholar] [PubMed] [CrossRef]

15. Ansper A, Alikas K. Retrieval of chlorophyll a from Sentinel-2 MSI data for the European union water framework directive reporting purposes. Remote Sens. 2019;11(1):64. doi:10.3390/rs11010064. [Google Scholar] [CrossRef]

16. Mansaray AS, Dzialowski AR, Martin ME, Wagner KL, Gholizadeh H, Stoodley SH. Comparing PlanetScope to Landsat-8 and Sentinel-2 for sensing water quality in reservoirs in agricultural watersheds. Remote Sens. 2021;13(9):1847. doi:10.3390/rs13091847. [Google Scholar] [CrossRef]

17. Zhang Z, Zhu L. A review on unmanned aerial vehicle remote sensing: platforms, sensors, data processing methods, and applications. Drones. 2023;7(6):398. doi:10.3390/drones7060398. [Google Scholar] [CrossRef]

18. Yang Z, Gong C, Lu Z, Wu E, Huai H, Hu Y, et al. Combined retrievals of turbidity from Sentinel-2A/B and landsat-8/9 in the Taihu Lake through machine learning. Remote Sens. 2023;15(17):4333. doi:10.3390/rs15174333. [Google Scholar] [CrossRef]

19. Bao Pham Q, Ajim Ali S, Parvin F, Van On V, Mohd Sidek L, Đurin B, et al. Multi-spectral remote sensing and GIS-based analysis for decadal land use land cover changes and future prediction using random forest tree and artificial neural network. Adv Space Res. 2024;74(1):17–47. doi:10.1016/j.asr.2024.03.027. [Google Scholar] [CrossRef]

20. Jeihouni M, Toomanian A, Alavipanah SK, Hamzeh S, Pilesjö P. Long term groundwater balance and water quality monitoring in the eastern Plains of Urmia Lake, Iran: a novel GIS based low cost approach. J Afr Earth Sci. 2018;147(1):11–9. doi:10.1016/j.jafrearsci.2018.06.017. [Google Scholar] [CrossRef]

21. Lemenkova P. Machine learning algorithms of remote sensing data processing for mapping changes in land cover types over central Apennines, Italy. J Imag. 2025;11(5):153. doi:10.3390/jimaging11050153. [Google Scholar] [PubMed] [CrossRef]

22. Oseke FI, Anornu GK, Adjei KA, Eduvie MO. Assessment of water quality using GIS techniques and water quality index in reservoirs affected by water diversion. Water Energy Nexus. 2021;4(1):25–34. doi:10.1016/j.wen.2020.12.002. [Google Scholar] [CrossRef]

23. Das A. Evaluation and prediction of surface water quality status for drinking purposes using an integrated water quality indices, GIS approaches, and machine learning techniques. Desalin Water Treat. 2025;323(1):101350. doi:10.1016/j.dwt.2025.101350. [Google Scholar] [CrossRef]

24. Shukla BK, Gupta L, Parashar B, Sharma PK, Sihag P, Shukla AK. Integrative assessment of surface water contamination using GIS, WQI, and machine learning in Urban–industrial confluence zones surrounding the national capital territory of the republic of India. Water. 2025;17(7):1076. doi:10.3390/w17071076. [Google Scholar] [CrossRef]

25. Kefi M, Aden MM, Ben Ali B. Water quality monitoring for irrigation by the integration of water quality index in a geographic information system environment in Chiba watershed, Nabeul, Tunisia. Water Conserv Sci Eng. 2024;9(1):20. doi:10.1007/s41101-024-00254-y. [Google Scholar] [CrossRef]

26. Gupta S, Gupta SK. A critical review on water quality index tool: genesis, evolution and future directions. Ecol Inform. 2021;63(17):101299. doi:10.1016/j.ecoinf.2021.101299. [Google Scholar] [CrossRef]

27. Sagan V, Peterson KT, Maimaitijiang M, Sidike P, Sloan J, Greeling BA, et al. Monitoring inland water quality using remote sensing: potential and limitations of spectral indices, bio-optical simulations, machine learning, and cloud computing. Earth Sci Rev. 2020;205(3):103187. doi:10.1016/j.earscirev.2020.103187. [Google Scholar] [CrossRef]

28. Muhammad A, Shangguan D, Rasool G, Ali Khan A, Butt AQ, Hussain A, et al. A localized evaluation of surface water quality using GIS-based water quality index along Satpara Watershed Skardu Baltistan, Pakistan. ISPRS Int J Geo Inf. 2024;13(11):393. doi:10.3390/ijgi13110393. [Google Scholar] [CrossRef]

29. Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. Int J Surg. 2010;8(5):336–41. doi:10.1016/j.ijsu.2010.02.007. [Google Scholar] [PubMed] [CrossRef]

30. World Bank. World bank group data 2026 [Internet]. 2026 [cited 2026 Feb 9]. Available from: https://data.worldbank.org/country. [Google Scholar]

31. Al-Shaibah B, Liu X, Zhang J, Tong Z, Zhang M, El-Zeiny A, et al. Modeling water quality parameters using landsat multispectral images: a case study of Erlong Lake, Northeast China. Remote Sens. 2021;13(9):1603. doi:10.3390/rs13091603. [Google Scholar] [CrossRef]

32. Gacu JG, Monjardin CEF, Mangulabnan RGT, Pugat GCE, Solmerin JG. Artificial intelligence (AI) in surface water management: a comprehensive review of methods, applications, and challenges. Water. 2025;17(11):1707. doi:10.3390/w17111707. [Google Scholar] [CrossRef]

33. Mikelsons K, Wang M, Wang XL, Jiang L. Global land mask for satellite ocean color remote sensing. Remote Sens Environ. 2021;257:112356. doi:10.1016/j.rse.2021.112356. [Google Scholar] [CrossRef]

34. Sudduth KA, Jang GS, Lerch RN, Sadler EJ. Long-term agroecosystem research in the central Mississippi river basin: hyperspectral remote sensing of reservoir water quality. J Environ Qual. 2015;44(1):71–83. doi:10.2134/jeq2014.02.0060. [Google Scholar] [PubMed] [CrossRef]

35. Chen J, Quan W. Using Landsat/TM imagery to estimate nitrogen and phosphorus concentration in Taihu Lake, China. IEEE J Sel Top Appl Earth Obs Remote Sens. 2012;5(1):273–80. doi:10.1109/JSTARS.2011.2174339. [Google Scholar] [PubMed] [CrossRef]

36. Niu C, Tan K, Jia X, Wang X. Deep learning based regression for optically inactive inland water quality parameter estimation using airborne hyperspectral imagery. Environ Pollut. 2021;286:117534. doi:10.1016/j.envpol.2021.117534. [Google Scholar] [PubMed] [CrossRef]

37. Maimouni S, Moufkari AA, Daghor L, Fekri A, Oubraim S, Lhissou R. Spatiotemporal monitoring of low water turbidity in Moroccan coastal lagoon using Sentinel-2 data. Remote Sens Appl Soc Environ. 2022;26(4):100772. doi:10.1016/j.rsase.2022.100772. [Google Scholar] [CrossRef]

38. Muchini R, Gumindoga W, Togarepi S, Masarira TP, Dube T. Near real time water quality monitoring of Chivero and Manyame lakes of Zimbabwe. Proc IAHS. 2018;378:85–92. doi:10.5194/piahs-378-85-2018. [Google Scholar] [CrossRef]

39. Shan J, Toth CK. Topographic laser ranging and scanning: principles and processing: principles and processing. 2nd ed. Boca Raton, FL, USA: CRC Press; 2018. doi:10.1201/9781315154381. [Google Scholar] [CrossRef]

40. Cai X, Wu L, Li Y, Lei S, Xu J, Lyu H, et al. Remote sensing identification of urban water pollution source types using hyperspectral data. J Hazard Mater. 2023;459:132080. doi:10.1016/j.jhazmat.2023.132080. [Google Scholar] [PubMed] [CrossRef]

41. Yang Z, Yu X, Dedman S, Rosso M, Zhu J, Yang J, et al. UAV remote sensing applications in marine monitoring: knowledge visualization and review. Sci Total Environ. 2022;838(Pt 1):155939. doi:10.1016/j.scitotenv.2022.155939. [Google Scholar] [PubMed] [CrossRef]

42. Wang M, Zhou C, Shi J, Lin F, Li Y, Hu Y, et al. Inversion of water quality parameters from UAV hyperspectral data based on intelligent algorithm optimized backpropagation neural networks of a small rural river. Remote Sens. 2025;17(1):119. doi:10.3390/rs17010119. [Google Scholar] [CrossRef]

43. Hu W, Liu J, Wang H, Miao D, Shao D, Gu W. Retrieval of TP concentration from UAV multispectral images using IOA-ML models in small inland waterbodies. Remote Sens. 2023;15(5):1250. doi:10.3390/rs15051250. [Google Scholar] [CrossRef]

44. Kim E, Nam SH, Ahn CH, Lee S, Koo JW, Hwang TM. Comparison of spatial interpolation methods for distribution map an unmanned surface vehicle data for chlorophyll-a monitoring in the stream. Environ Technol Innov. 2022;28(3.11):102637. doi:10.1016/j.eti.2022.102637. [Google Scholar] [CrossRef]

45. Mohan S, Kumar B, Nejadhashemi AP. Integration of machine learning and remote sensing for water quality monitoring and prediction: a review. Sustainability. 2025;17(3):998. doi:10.3390/su17030998; [Google Scholar] [CrossRef]

46. Cao X, Zhang J, Meng H, Lai Y, Xu M. Remote sensing inversion of water quality parameters in the Yellow River Delta. Ecol Indic. 2023;155(10):110914. doi:10.1016/j.ecolind.2023.110914; [Google Scholar] [CrossRef]

47. Adilakshmi A, Venkatesan V. Effective monitoring of Noyyal River surface water quality using remote sensing and machine learning and GIS techniques. Desalin Water Treat. 2024;320(6):100630. doi:10.1016/j.dwt.2024.100630. [Google Scholar] [CrossRef]

48. Ahmed W, Mohammed S, El-Shazly A, Morsy S. Tigris River water surface quality monitoring using remote sensing data and GIS techniques. Egypt J Remote Sens Space Sci. 2023;26(3):816–25. doi:10.1016/j.ejrs.2023.09.001. [Google Scholar] [CrossRef]

49. Ahmad I, Alasgah AA, Zelenakova M, Ahmad Dar M, Damtie M, Berhan M. Quantifying turbidity dynamics in lake water using OLS regression: a Landsat 8 OLI-based remote sensing approach. J Hydrol Reg Stud. 2025;60(1):102523. doi:10.1016/j.ejrh.2025.102523. [Google Scholar] [CrossRef]

50. Assegide E, Shiferaw H, Tibebe D, Peppa MV, Walsh CL, Alamirew T, et al. Spatiotemporal dynamics of water quality indicators in Koka Reservoir, Ethiopia. Remote Sens. 2023;15(4):1155. doi:10.3390/rs15041155; [Google Scholar] [CrossRef]

51. Najafzadeh M, Basirian S. Evaluation of river water quality index using remote sensing and artificial intelligence models. Remote Sens. 2023;15(9):2359. doi:10.3390/rs15092359. [Google Scholar] [CrossRef]

52. Jakovljevic G, Álvarez-Taboada F, Govedarica M. Long-term monitoring of inland water quality parameters using Landsat time-series and back-propagated ANN: assessment and usability in a real-case scenario. Remote Sens. 2024;16(1):68. doi:10.3390/rs16010068. [Google Scholar] [CrossRef]

53. Eljaiek-Urzola M, Sander de Carvalho LA, Betancur-Turizo SP, Quiñones-Bolaños E, Castrillón-Ortiz C. Spatial patterns of turbidity in Cartagena Bay, Colombia, using Sentinel-2 imagery. Remote Sens. 2024;16(1):179. doi:10.3390/rs16010179; [Google Scholar] [CrossRef]

54. Villota-González FH, Sulbarán-Rangel B, Zurita-Martínez F, Gurubel-Tun KJ, Zúñiga-Grajeda V. Assessment of machine learning models for remote sensing of water quality in Lakes Cajititlán and Zapotlán, Jalisco—Mexico. Remote Sens. 2023;15(23):5505. doi:10.3390/rs15235505; [Google Scholar] [CrossRef]

55. Rodríguez-López L, Usta DB, Duran-Llacer I, Alvarez LB, Yépez S, Bourrel L, et al. Estimation of water quality parameters through a combination of deep learning and remote sensing techniques in a lake in southern Chile. Remote Sens. 2023;15(17):4157. doi:10.3390/rs15174157. [Google Scholar] [CrossRef]

56. Zhao Y, He X, Pan S, Bai Y, Wang D, Li T, et al. Satellite retrievals of water quality for diverse inland waters from Sentinel-2 images: an example from Zhejiang Province, China. Int J Appl Earth Obs Geoinf. 2024;132:104048. doi:10.1016/j.jag.2024.104048. [Google Scholar] [CrossRef]

57. Zhou Y, Yang W, Hu M, Li J, Liu X. Research on remote sensing inversion of total phosphorus in east Juyan Lake based on machine learning. Hydrology. 2025;12(11):299. doi:10.3390/hydrology12110299; [Google Scholar] [CrossRef]

58. Sun Z, Guo L, Tao Z, Li Y, Zhan Y, Li S, et al. Water quality inversion framework for Taihu Lake based on multilayer denoising autoencoder and ensemble learning. Remote Sens. 2024;16(24):4793. doi:10.3390/rs16244793. [Google Scholar] [CrossRef]

59. Rodríguez-López L, Duran-Llacer I, Bravo Alvarez L, Lami A, Urrutia R. Recovery of water quality and detection of algal blooms in Lake Villarrica through Landsat satellite images and monitoring data. Remote Sens. 2023;15(7):1929. doi:10.3390/rs15071929. [Google Scholar] [CrossRef]

60. Wang Z, Mao Z, Zhang L, Zhang X, Yuan D, Li Y, et al. Observations of the impacts of Hong Kong international airport on water quality from 1986 to 2022 using Landsat satellite. Remote Sens. 2023;15(12):3146. doi:10.3390/rs15123146. [Google Scholar] [CrossRef]

61. Jin H, Fang S, Chen C. Mapping of the spatial scope and water quality of surface water based on the google earth engine cloud platform and Landsat time series. Remote Sens. 2023;15(20):4986. doi:10.3390/rs15204986. [Google Scholar] [CrossRef]

62. Kiwanuka M, Leslie R, Gidudu A, Obubu JP, Melesse A, Balaji Bhaskar MS. Evaluating eutrophication and water clarity on lake Victoria’s Ugandan coast using landsat data. Sustainability. 2025;17(20):9056. doi:10.3390/su17209056. [Google Scholar] [CrossRef]

63. Kiwanuka M, Oyege I, Balaji Bhaskar MS. Spatial and temporal dynamics of water quality in Lake Okeechobee using remote sensing and its impact on environmental health. Remote Sens. 2025;17(18):3197. doi:10.3390/rs17183197. [Google Scholar] [CrossRef]

64. Schmidt SI, Schröder T, Kutzner RD, Laue P, Bernert H, Stelzer K, et al. Evaluating satellite-based water quality sensing of inland waters on basis of 100+ German water bodies using 2 different processing chains. Remote Sens. 2024;16(18):3416. doi:10.3390/rs16183416. [Google Scholar] [CrossRef]

65. Rivera-Ruiz D, Arumí JL, Lillo-Saavedra M, Esse C, Arancibia-Ávila P, Urrutia R, et al. Secchi depth retrieval in oligotrophic to eutrophic Chilean Lakes using open access satellite-derived products. Remote Sens. 2024;16(22):4327. doi:10.3390/rs16224327. [Google Scholar] [CrossRef]

66. Qing S, Cui T, Lai Q, Bao Y, Diao R, Yue Y, et al. Improving remote sensing retrieval of water clarity in complex coastal and inland waters with modified absorption estimation and optical water classification using Sentinel-2 MSI. Int J Appl Earth Obs Geoinf. 2021;102:102377. doi:10.1016/j.jag.2021.102377. [Google Scholar] [CrossRef]

67. Zhang Z, Li C, Yang P, Xu Z, Yao L, Wang Q, et al. Enhancing remote sensing water quality inversion through integration of multisource spatial covariates: a case study of Hong Kong’s coastal nutrient concentrations. Remote Sens. 2024;16(17):3337. doi:10.3390/rs16173337. [Google Scholar] [CrossRef]

68. Du Y, Zhang Y, Ling F, Wang Q, Li W, Li X. Water bodies’ mapping from Sentinel-2 imagery with modified normalized difference water index at 10-m spatial resolution produced by sharpening the SWIR band. Remote Sens. 2016;8(4):354. doi:10.3390/rs8040354. [Google Scholar] [CrossRef]

69. Valerio F, Godinho S, Ferraz G, Pita R, Gameiro J, Silva B, et al. Multi-temporal remote sensing of inland surface waters: a fusion of Sentinel-1&2 data applied to small seasonal ponds in semiarid environments. Int J Appl Earth Obs Geoinf. 2024;135(5):104283. doi:10.1016/j.jag.2024.104283. [Google Scholar] [CrossRef]

70. Choto M, Higa H, Salem SI, Siswanto E, Suzuki T, Mäll M. Performance evaluation of inherent optical property algorithms and identification of potential water quality indicators using GCOM-C data in Eutrophic Lake Kasumigaura, Japan. Remote Sens. 2025;17(9):1621. doi:10.3390/rs17091621. [Google Scholar] [CrossRef]

71. Gupta A, Hantush MM, Govindaraju RS. Sub-monthly time scale forecasting of harmful algal blooms intensity in Lake Erie using remote sensing and machine learning. Sci Total Environ. 2023;900(3):165781. doi:10.1016/j.scitotenv.2023.165781. [Google Scholar] [PubMed] [CrossRef]

72. Salem SI, Toratani M, Higa H, Son S, Siswanto E, Ishizaka J. Long-term evaluation of GCOM-C/SGLI reflectance and water quality products: variability among JAXA G-portal and JASMES. Remote Sens. 2025;17(2):221. doi:10.3390/rs17020221. [Google Scholar] [CrossRef]

73. Flores R Jr, Barbosa CCF, Maciel DA, de Moraes Novo EML, Martins VS, de Lucia Lobo F, et al. Hybrid semi-analytical algorithm for estimating chlorophyll-a concentration in lower Amazon floodplain waters. Front Remote Sens. 2022;3:834576. doi:10.3389/frsen.2022.834576. [Google Scholar] [CrossRef]

74. Binding CE, Pizzolato L, Zeng C. EOLakeWatch; delivering a comprehensive suite of remote sensing algal bloom indices for enhanced monitoring of Canadian Eutrophic Lakes. Ecol Indic. 2021;121(4):106999. doi:10.1016/j.ecolind.2020.106999. [Google Scholar] [CrossRef]

75. Kim YH, Son S, Kim HC, Kim B, Park YG, Nam J, et al. Application of satellite remote sensing in monitoring dissolved oxygen variabilities: a case study for coastal waters in Korea. Environ Int. 2020;134:105301. doi:10.1016/j.envint.2019.105301. [Google Scholar] [PubMed] [CrossRef]

76. Alharbi B. Remote sensing techniques for monitoring algal blooms in the area between Jeddah and Rabigh on the Red Sea Coast. Remote Sens Appl Soc Environ. 2023;30(10):100935. doi:10.1016/j.rsase.2023.100935. [Google Scholar] [CrossRef]

77. Caballero I, Román A, Tovar-Sánchez A, Navarro G. Water quality monitoring with Sentinel-2 and Landsat-8 satellites during the 2021 volcanic eruption in La Palma (Canary Islands). Sci Total Environ. 2022;822(SI):153433. doi:10.1016/j.scitotenv.2022.153433. [Google Scholar] [PubMed] [CrossRef]

78. Zhong A, Wang D, Gong F, Zhu W, Fu D, Zheng Z, et al. Remote sensing estimates of global sea surface nitrate: methodology and validation. Sci Total Environ. 2024;950(4):175362. doi:10.1016/j.scitotenv.2024.175362. [Google Scholar] [PubMed] [CrossRef]

79. Dawn A, Hinge G, Kumar A, Nikoo MR, Hamouda MA. Assessment of water quality in urban lakes using multi-source data and modeling techniques. Sustainability. 2025;17(16):7258. doi:10.3390/su17167258. [Google Scholar] [CrossRef]

80. Coccia C, Pintado E, Paredes ÁL, Aragonés D, O’Ryan DC, Green AJ, et al. Modelling water depth, turbidity and chlorophyll using airborne hyperspectral remote sensing in a restored pond complex of Doñana National Park (Spain). Remote Sens. 2024;16(16):2996. doi:10.3390/rs16162996. [Google Scholar] [CrossRef]

81. Lei X, Jiang J, Deng Z, Wu D, Wang F, Lai C, et al. An ensemble machine learning model to estimate urban water quality parameters using unmanned aerial vehicle multispectral imagery. Remote Sens. 2024;16(12):2246. doi:10.3390/rs16122246. [Google Scholar] [CrossRef]

82. Luo X, Li N, Zhang Y, Zhang Y, Shi K, Qin B, et al. High-frequency monitoring and short-term forecasting of surface water temperature using a novel hyperspectral proximal sensing system. Remote Sens. 2025;17(19):3303. doi:10.3390/rs17193303. [Google Scholar] [CrossRef]

83. Zhang F, Duan P, Jim CY, Johnson VC, Liu C, Chan NW, et al. An advanced spatiotemporal fusion model for suspended particulate matter monitoring in an Intermontane Lake. Remote Sens. 2023;15(5):1204. doi:10.3390/rs15051204. [Google Scholar] [CrossRef]

84. Fickas KC, O’Shea RE, Pahlevan N, Smith B, Bartlett SL, Wolny JL. Leveraging multimission satellite data for spatiotemporally coherent cyanoHAB monitoring. Front Remote Sens. 2023;4:1157609. doi:10.3389/frsen.2023.1157609; [Google Scholar] [CrossRef]

85. Igoe A, Möller I, Basu B. Modelling suspended sediment concentration in coastal Ireland using machine learning. Camb Prism Coast Futur. 2025;3:e26. doi:10.1017/cft.2025.10016. [Google Scholar] [PubMed] [CrossRef]

86. Al Mamun Hridoy MA, Paul PB, Masood A. Monitoring flood-prone Urban areas of Sylhet, Bangladesh through water quality remote sensing. Discov Environ. 2025;3(1):125. doi:10.1007/s44274-025-00322-7. [Google Scholar] [CrossRef]

87. Bygate M, Ahmed M. Monitoring water quality indicators over Matagorda Bay, Texas, using Landsat-8. Remote Sens. 2024;16(7):1120. doi:10.3390/rs16071120. [Google Scholar] [CrossRef]

88. Sang X, Mao Z, Li Y, Zhang X, Han C, Zhang L, et al. Observations of optical properties and chlorophyll-a concentration in Qiandao Lake using shipborne lidar. Remote Sens. 2024;16(24):4663. doi:10.3390/rs16244663; [Google Scholar] [CrossRef]

89. Pelevin V, Koltsova E, Molkov A, Fedorov S, Alymkulov S, Konovalov B, et al. Regional models for Sentinel-2/MSI imagery of chlorophyll a and TSS, obtained for oligotrophic Issyk-Kul lake using high-resolution LIF LiDAR data. Remote Sens. 2023;15(18):4443. doi:10.3390/rs15184443. [Google Scholar] [CrossRef]

90. Jiang D, Jones I, Liu X, Simis SGH, Cretaux JF, Albergel C, et al. Impacts of droughts and human activities on water quantity and quality: remote sensing observations of Lake Qadisiyah, Iraq. Int J Appl Earth Obs Geoinf. 2024;132:104021. doi:10.1016/j.jag.2024.104021. [Google Scholar] [CrossRef]

91. Kharabsheh RM, Bdour AN. Predictive modeling coupled with wireless sensor networks for sustainable marine ecosystem management using real-time remote monitoring of water quality. Open Eng. 2025;15(1):20250129. doi:10.1515/eng-2025-0129. [Google Scholar] [PubMed] [CrossRef]

92. Xu H, Xu G, Wen X, Hu X, Wang Y. Lockdown effects on total suspended solids concentrations in the Lower Min River (China) during COVID-19 using time-series remote sensing images. Int J Appl Earth Obs Geoinf. 2021;98(1):102301. doi:10.1016/j.jag.2021.102301. [Google Scholar] [PubMed] [CrossRef]

93. Ali A, Zhou G, Pablo Antezana Lopez F, Xu C, Jing G, Tan Y. Deep learning for water quality multivariate assessment in inland water across China. Int J Appl Earth Obs Geoinf. 2024;133(4):104078. doi:10.1016/j.jag.2024.104078. [Google Scholar] [CrossRef]

94. Joshi N, Park J, Zhao K, Londo A, Khanal S. Monitoring harmful algal blooms and water quality using Sentinel-3 OLCI satellite imagery with machine learning. Remote Sens. 2024;16(13):2444. doi:10.3390/rs16132444. [Google Scholar] [CrossRef]

95. Li P, Hao F, Wu H, Nie H. Spatiotemporal dynamic analysis of eutrophication status based on machine learning-based retrieval algorithm: case study in Liangzi Lake, Hubei, China. Remote Sens. 2024;16(22):4192. doi:10.3390/rs16224192. [Google Scholar] [CrossRef]

96. Rao K, Cao X, Wang Y, Zhang Y, Huang H, Ma Y, et al. Spatial-temporal distributions of phytoplankton shifting, chlorophyll-a, and their influencing factors in shallow lakes using remote sensing. Ecol Inform. 2024;82(1):102765. doi:10.1016/j.ecoinf.2024.102765; [Google Scholar] [CrossRef]

97. Zhang X, Huang J, Chen J, Zhao Y. Remote sensing monitoring of total suspended solids concentration in Jiaozhou Bay based on multi-source data. Ecol Indic. 2023;154(1):110513. doi:10.1016/j.ecolind.2023.110513. [Google Scholar] [CrossRef]

98. Dong L, Wang D, Song L, Gong F, Chen S, Huang J, et al. Monitoring dissolved oxygen concentrations in the coastal waters of Zhejiang using Landsat-8/9 imagery. Remote Sens. 2024;16(11):1951. doi:10.3390/rs16111951. [Google Scholar] [CrossRef]

99. Hong SM, Morgan BJ, Stocker MD, Smith JE, Kim MS, Cho KH, et al. Using machine learning models to estimate Escherichia coli concentration in an irrigation pond from water quality and drone-based RGB imagery data. Water Res. 2024;260(3):121861. doi:10.1016/j.watres.2024.121861. [Google Scholar] [PubMed] [CrossRef]

100. Kokolakis S, Kokinou E, Karagiannidou M, Gerarchakis N, Vasilakos C, Kotti M, et al. From space to stream: combining remote sensing and in situ techniques for comprehensive stream health assessment. Remote Sens. 2025;17(9):1532. doi:10.3390/rs17091532; [Google Scholar] [CrossRef]

101. Li H, Wang N, Du Z, Huang D, Shi M, Zhong Z, et al. Multi-parameter water quality inversion in heterogeneous inland waters using UAV-based hyperspectral data and deep learning methods. Remote Sens. 2025;17(13):2191. doi:10.3390/rs17132191. [Google Scholar] [CrossRef]

102. Bi S, Li Y, Wang Q, Lyu H, Liu G, Zheng Z, et al. Inland water atmospheric correction based on turbidity classification using OLCI and SLSTR synergistic observations. Remote Sens. 2018;10(7):1002. doi:10.3390/rs10071002. [Google Scholar] [CrossRef]

103. Paulino RS, Martins VS, Novo EMLM, Barbosa CCF, de Carvalho LAS, Begliomini FN. Assessment of adjacency correction over inland waters using Sentinel-2 MSI images. Remote Sens. 2022;14(8):1829. doi:10.3390/rs14081829. [Google Scholar] [CrossRef]

104. Martins VS, Barbosa CCF, De Carvalho LAS, Jorge DSF, De Lucia Lobo F, de Moraes Novo EML. Assessment of atmospheric correction methods for Sentinel-2 MSI images applied to Amazon floodplain lakes. Remote Sens. 2017;9(4):322. doi:10.3390/rs9040322. [Google Scholar] [CrossRef]

105. Adjovu GE, Stephen H, James D, Ahmad S. Overview of the application of remote sensing in effective monitoring of water quality parameters. Remote Sens. 2023;15(7):1938. doi:10.3390/rs15071938. [Google Scholar] [CrossRef]

106. Gholizadeh MH, Melesse AM, Reddi L. A comprehensive review on water quality parameters estimation using remote sensing techniques. Sensors. 2016;16(8):1298. doi:10.3390/s16081298. [Google Scholar] [PubMed] [CrossRef]

107. Dey J, Vijay R. A critical and intensive review on assessment of water quality parameters through geospatial techniques. Environ Sci Pollut Res Int. 2021;28(31):41612–26. doi:10.1007/s11356-021-14726-4; [Google Scholar] [CrossRef]

108. Kowe P, Ncube E, Magidi J, Ndambuki JM, Rwasoka DT, Gumindoga W, et al. Spatial-temporal variability analysis of water quality using remote sensing data: a case study of Lake Manyame. Sci Afr. 2023;21(1):e01877. doi:10.1016/j.sciaf.2023.e01877. [Google Scholar] [CrossRef]

109. Kandasamy L, Mahendran A, Sangaraju SHV, Mathur P, Faldu SV, Mazzara M. Enhanced remote sensing and deep learning aided water quality detection in the Ganges River, India supporting monitoring of aquatic environments. Results Eng. 2025;25(3):103604. doi:10.1016/j.rineng.2024.103604. [Google Scholar] [CrossRef]

110. Tavares MH, Guimarães D, Roussillon J, Baute V, Cucherousset J, Boulêtreau S, et al. A framework to retrieve water quality parameters in small, optically diverse freshwater ecosystems using Sentinel-2 MSI imagery. Remote Sens. 2025;17(15):2729. doi:10.3390/rs17152729; [Google Scholar] [CrossRef]

111. Assaf MN, Abdelal Q, Hussein NM, Halaweh G, Alzubaidi AJ. Water quality monitoring and management: integration of machine learning algorithms and Sentinel-2 images for the estimation of chlorophyll-a. Model Earth Syst Environ. 2025;11(5):348. doi:10.1007/s40808-025-02543-4. [Google Scholar] [CrossRef]

112. Chen J, Feng X, Guo C, Chen Y, Tong F, Zhang L, et al. Assessing the impacts of marine ranching construction on water quality and fishery resources in adjacent coastal waters. Remote Sens. 2025;17(18):3140. doi:10.3390/rs17183140. [Google Scholar] [CrossRef]

113. Steinbach S, Bartels A, Rienow A, Thiong’o Kuria B, Zwart SJ, Nelson A. Predicting turbidity dynamics in small reservoirs in central Kenya using remote sensing and machine learning. Int J Appl Earth Obs Geoinf. 2025;136:104390. doi:10.1016/j.jag.2025.104390. [Google Scholar] [CrossRef]

114. Kong Y, Jimenez K, Lee CM, Winter S, Summers-Evans J, Cao A, et al. Monitoring coastal water turbidity using Sentinel2—a case study in Los Angeles. Remote Sens. 2025;17(2):201. doi:10.3390/rs17020201. [Google Scholar] [CrossRef]

115. Patricio-Valerio L, Schroeder T, Devlin MJ, Qin Y, Smithers S. Meteorological satellite observations reveal diurnal exceedance of water quality guideline thresholds in the coastal great barrier reef. Remote Sens. 2023;15(9):2335. doi:10.3390/rs15092335; [Google Scholar] [CrossRef]

116. Maniyar CB, Rudresh M, Callejas IA, Osborn K, Lee CM, Jay J, et al. Spatio-temporal dynamics of total suspended sediments in the Belize coastal lagoon. Remote Sens. 2023;15(23):5625. doi:10.3390/rs15235625; [Google Scholar] [CrossRef]

117. Chen X, Lou G, Li H, Zhang X, Liu S, Gao Q, et al. An enhanced machine learning approach for regional total suspended matter concentration retrieval using multispectral imagery. Water. 2025;17(22):3252. doi:10.3390/w17223252. [Google Scholar] [CrossRef]

118. Zhang H, Yao B, Wang S, Wang G. Remote sensing estimation of the concentration and sources of coloured dissolved organic matter based on MODIS: a case study of Erhai Lake. Ecol Indic. 2021;131(1):108180. doi:10.1016/j.ecolind.2021.108180. [Google Scholar] [CrossRef]

119. Dong L, Gong C, Huai H, Wu E, Lu Z, Hu Y, et al. Retrieval of water quality parameters in Dianshan Lake based on Sentinel-2 MSI imagery and machine learning: algorithm evaluation and spatiotemporal change research. Remote Sens. 2023;15(20):5001. doi:10.3390/rs15205001. [Google Scholar] [CrossRef]

120. Fu B, Lao Z, Liang Y, Sun J, He X, Deng T, et al. Evaluating optically and non-optically active water quality and its response relationship to hydro-meteorology using multi-source data in Poyang Lake, China. Ecol Indic. 2022;145:109675. doi:10.1016/j.ecolind.2022.109675. [Google Scholar] [CrossRef]

121. Arias-Rodriguez LF, Tüzün UF, Duan Z, Huang J, Tuo Y, Disse M. Global water quality of inland waters with harmonized Landsat-8 and Sentinel-2 using cloud-computed machine learning. Remote Sens. 2023;15(5):1390. doi:10.3390/rs15051390. [Google Scholar] [CrossRef]

122. Hu W, Jin S, Zhang Y. Water quality variations in the lower Yangtze River based on GA-RF model from GF-1, Landsat-8, and Sentinel-2 images. IEEE J Sel Top Appl Earth Obs Remote Sens. 2025;18(1):4992–5004. doi:10.1109/JSTARS.2025.3526207. [Google Scholar] [PubMed] [CrossRef]

123. Keith DJ. Satellite remote sensing of chlorophyll a in support of nutrient management in the Neuse and Tar–Pamlico River (North Carolina) estuaries. Remote Sens Environ. 2014;153(1):61–78. doi:10.1016/j.rse.2014.05.019. [Google Scholar] [CrossRef]

124. Muhoyi H, Gumindoga W, Mhizha A, Misi SN, Nondo N. Remote sensing application in compliment to in-situ monitoring of water quality: lower Manyame Sub-catchment, Zimbabwe. Sci Afr. 2025;27:e02551. doi:10.1016/j.sciaf.2025.e02551. [Google Scholar] [CrossRef]

125. Li H, Somogyi B, Chen X, Luo Z, Blix K, Wu S, et al. Leveraging Landsat and google earth engine for long-term chlorophyll-a monitoring: a case study of Lake Balaton’s water quality. Ecol Inform. 2025;90(9):103245. doi:10.1016/j.ecoinf.2025.103245. [Google Scholar] [CrossRef]

126. Li Z, Chen C, Cao N, Jiang Z, Liu C, Adeyinka Oke S, et al. High spatial resolution inversion of chromophoric dissolved organic matter (CDOM) concentrations in Ebinur Lake of arid Xinjiang, China: implications for surface water quality monitoring. Int J Appl Earth Obs Geoinf. 2024;132(6):104022. doi:10.1016/j.jag.2024.104022. [Google Scholar] [CrossRef]

127. Qiang S, Song K, Shang Y, Lai F, Wen Z, Liu G, et al. Remote sensing estimation of CDOM and DOC with the environmental implications for Lake Khanka. Remote Sens. 2023;15(24):5707. doi:10.3390/rs15245707. [Google Scholar] [CrossRef]

128. Yan Z, Fang C, Song K, Wang X, Wen Z, Shang Y, et al. Spatiotemporal variation in biomass abundance of different algal species in Lake Hulun using machine learning and Sentinel-3 images. Sci Rep. 2025;15(1):2739. doi:10.1038/s41598-025-87338-4. [Google Scholar] [PubMed] [CrossRef]

129. Barkhordar H, Mohammadpour G, Hassanzadeh S, Karemi H. Investigation of water quality in the shallow coastal waters of the Persian Gulf. Mar Pollut Bull. 2024;209(Pt B):117263. doi:10.1016/j.marpolbul.2024.117263. [Google Scholar] [PubMed] [CrossRef]

130. Joshi N, Ghoorkhanian A, Park J, Zhao K, Khanal S. A machine learning-based assessment of proxies and drivers of harmful algal blooms in the western Lake Erie Basin using satellite remote sensing. Remote Sens. 2025;17(13):2164. doi:10.3390/rs17132164. [Google Scholar] [CrossRef]

131. Anand V, Oinam B, Wieprecht S. Machine learning approach for water quality predictions based on multispectral satellite imageries. Ecol Inform. 2024;84:102868. doi:10.1016/j.ecoinf.2024.102868. [Google Scholar] [CrossRef]

132. Sedighkia M, Datta B, Saeedipour P, Abdoli A. Predicting water quality distribution of lakes through linking remote sensing-based monitoring and machine learning simulation. Remote Sens. 2023;15(13):3302. doi:10.3390/rs15133302. [Google Scholar] [CrossRef]

133. Mamun M, Hasan M, An KG. Advancing reservoirs water quality parameters estimation using Sentinel-2 and Landsat-8 satellite data with machine learning approaches. Ecol Inform. 2024;81:102608. doi:10.1016/j.ecoinf.2024.102608. [Google Scholar] [CrossRef]

134. Meng H, Zhang J, Zheng Z, Song Y, Lai Y. Classification of inland lake water quality levels based on Sentinel-2 images using convolutional neural networks and spatiotemporal variation and driving factors of algal bloom. Ecol Inform. 2024;80:102549. doi:10.1016/j.ecoinf.2024.102549. [Google Scholar] [CrossRef]

135. Xu J, Pan J, Devlin AT. Variations in chlorophyll-a concentration in response to hydrodynamics in a flow-through lake: remote sensing and modeling studies. Ecol Indic. 2023;148(1):110128. doi:10.1016/j.ecolind.2023.110128. [Google Scholar] [CrossRef]

136. Werther M, Odermatt D, Simis SGH, Gurlin D, Lehmann MK, Kutser T, et al. A Bayesian approach for remote sensing of chlorophyll-a and associated retrieval uncertainty in oligotrophic and Mesotrophic lakes. Remote Sens Environ. 2022;283(3):113295. doi:10.1016/j.rse.2022.113295. [Google Scholar] [CrossRef]

137. Li C, Odermatt D, Bouffard D, Wüest A, Kohn T. Coupling remote sensing and particle tracking to estimate trajectories in large water bodies. Intl J Appl Earth Obs Geoinf. 2022;110(8):102809. doi:10.1016/j.jag.2022.102809. [Google Scholar] [CrossRef]

138. Synan HE, Howes BL, Sampieri S, Lohrenz SE. Water quality monitoring using Landsat 8 OLI in pleasant bay, Massachusetts, USA. Remote Sens. 2025;17(4):638. doi:10.3390/rs17040638. [Google Scholar] [CrossRef]

139. Li X, Zhao Y, Shi C, Sha J, Wang ZL, Wang Y. Application of water evaluation and planning (WEAP) model for water resources management strategy estimation in coastal Binhai new area, China. Ocean Coast Manag. 2015;106:97–109. doi:10.1016/j.ocecoaman.2015.01.016. [Google Scholar] [CrossRef]

140. Ahmadi M, Noori A, Mohajeri SH, Nikoo MR. Integrating river discharge and Sentinel-2 satellite imagery for enhanced turbidity mapping in arid region rivers: a machine learning approach. Phys Chem Earth. 2025;138(13):103869. doi:10.1016/j.pce.2025.103869. [Google Scholar] [CrossRef]

141. Alruwais N, Marzouk R, Albalawneh D, Arasi MA, Shobana M, Kavitha R. Impact analysis of polluted waste water discharge in river and management process using machine learning and GIS approach. Desalin Water Treat. 2025;323(6):101323. doi:10.1016/j.dwt.2025.101323. [Google Scholar] [CrossRef]

142. Huang Z, Liu X, Sun S, Tang Y, Yuan X, Tang Q. Global assessment of future sectoral water scarcity under adaptive inner-basin water allocation measures. Sci Total Environ. 2021;783(7578):146973. doi:10.1016/j.scitotenv.2021.146973. [Google Scholar] [PubMed] [CrossRef]

143. Sterckx S, Knaeps S, Kratzer S, Ruddick K. SIMilarity Environment Correction (SIMEC) applied to MERIS data over inland and coastal waters. Remote Sens Environ. 2015;157(9):96–110. doi:10.1016/j.rse.2014.06.017. [Google Scholar] [CrossRef]

144. Liu Y, Xiao C, Li J, Zhang F, Wang S. Secchi disk depth estimation from China’s new generation of GF-5 hyperspectral observations using a semi-analytical scheme. Remote Sens. 2020;12(11):1849. doi:10.3390/rs12111849. [Google Scholar] [CrossRef]

145. Islam MT, Akash, Khatun MR, Jahan N, Islam MR, Partho DK, et al. Regional irrigation water quality index for the Old Brahmaputra River, Bangladesh: a multivariate and GIS-based spatiotemporal assessment. Results Eng. 2024;24(35):103667. doi:10.1016/j.rineng.2024.103667. [Google Scholar] [CrossRef]

146. Xiao Y, Chen J, Xu Y, Guo S, Nie X, Guo Y, et al. Monitoring of chlorophyll-a and suspended sediment concentrations in optically complex inland rivers using multisource remote sensing measurements. Ecol Indic. 2023;155:111041. doi:10.1016/j.ecolind.2023.111041. [Google Scholar] [CrossRef]

147. Chen L, Liu L, Liu S, Shi Z, Shi C. The application of remote sensing technology in inland water quality monitoring and water environment science: recent progress and perspectives. Remote Sens. 2025;17(4):667. doi:10.3390/rs17040667. [Google Scholar] [CrossRef]

148. Batina A, Šiljeg A. Enhancing water quality monitoring in a coastal shallow lake using GIS and multi-criteria decision analysis. Environ Sustain Indic. 2025;28:100881. doi:10.1016/j.indic.2025.100881; [Google Scholar] [CrossRef]

149. Mussina A, Aktymbayeva A, Zhanabayeva Z, Mashtayeva S, Macklin MG, Rysmagambetova A, et al. A decade-long assessment of water quality variability in the Yelek River basin (Kazakhstan) using remote sensing and GIS. Sustainability. 2025;17(21):9809. doi:10.3390/su17219809. [Google Scholar] [CrossRef]