literature review on river water quality

Literature review: Water quality and public health problems in developing countries

• Article contents
• Figures & tables
• Supplementary Data
• Peer Review
• Reprints and Permissions
• Cite Icon Cite
• Search Site

Eni Muryani; Literature review: Water quality and public health problems in developing countries. AIP Conf. Proc. 23 November 2021; 2363 (1): 050020. https://doi.org/10.1063/5.0061561

Download citation file:

• Ris (Zotero)
• Reference Manager

Water’s essential function as drinking water is a significant daily intake. Contamination by microorganisms (bacteria or viruses) on water sources and drinking water supplies is a common cause in developing countries like Indonesia. This paper will discuss the sources of clean water and drinking water and their problems in developing countries; water quality and its relation to public health problems in these countries; and what efforts that can be make to improve water quality. The method used is a literature review from the latest journals. Water quality is influenced by natural processes and human activities around the water source Among developed countries, public health problems caused by low water quality, such as diarrhea, dysentery, cholera, typhus, skin itching, kidney disease, hypertension, heart disease, cancer, and other diseases the nervous system. Good water quality has a role to play in decreasing the number of disease sufferers or health issues due to drinking and the mortality rate. The efforts made to improve water quality and public health are by improving WASH (water, sanitation, and hygiene) facilities and infrastructure and also WASH education.

Citing articles via

Publish with us - request a quote.

Sign up for alerts

• Online ISSN 1551-7616
• Print ISSN 0094-243X
• For Researchers
• For Librarians
• For Advertisers
• Our Publishing Partners  
• Physics Today
• Conference Proceedings
• Special Topics

pubs.aip.org

• Privacy Policy
• Terms of Use

Connect with AIP Publishing

This feature is available to subscribers only.

Sign In or Create an Account

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 29 March 2024

Reliable water quality prediction and parametric analysis using explainable AI models

• M. K. Nallakaruppan 1 ,
• E. Gangadevi 2 ,
• M. Lawanya Shri 1 ,
• Balamurugan Balusamy 3 ,
• Sweta Bhattacharya 1 &
• Shitharth Selvarajan 4 , 5  

Scientific Reports volume  14 , Article number:  7520 ( 2024 ) Cite this article

4594 Accesses

9 Citations

Metrics details

• Electrical and electronic engineering
• Engineering

The consumption of water constitutes the physical health of most of the living species and hence management of its purity and quality is extremely essential as contaminated water has to potential to create adverse health and environmental consequences. This creates the dire necessity to measure, control and monitor the quality of water. The primary contaminant present in water is Total Dissolved Solids (TDS), which is hard to filter out. There are various substances apart from mere solids such as potassium, sodium, chlorides, lead, nitrate, cadmium, arsenic and other pollutants. The proposed work aims to provide the automation of water quality estimation through Artificial Intelligence and uses Explainable Artificial Intelligence (XAI) for the explanation of the most significant parameters contributing towards the potability of water and the estimation of the impurities. XAI has the transparency and justifiability as a white-box model since the Machine Learning (ML) model is black-box and unable to describe the reasoning behind the ML classification. The proposed work uses various ML models such as Logistic Regression, Support Vector Machine (SVM), Gaussian Naive Bayes, Decision Tree (DT) and Random Forest (RF) to classify whether the water is drinkable. The various representations of XAI such as force plot, test patch, summary plot, dependency plot and decision plot generated in SHAPELY explainer explain the significant features, prediction score, feature importance and justification behind the water quality estimation. The RF classifier is selected for the explanation and yields optimum Accuracy and F1-Score of 0.9999, with Precision and Re-call of 0.9997 and 0.998 respectively. Thus, the work is an exploratory analysis of the estimation and management of water quality with indicators associated with their significance. This work is an emerging research at present with a vision of addressing the water quality for the future as well.

Similar content being viewed by others

A proficiency assessment of integrating machine learning (ML) schemes on Lahore water ensemble

Source discrimination of mine water based on the random forest method

Computational assessment of groundwater salinity distribution within coastal multi-aquifers of Bangladesh

Introduction.

The major part of our earth comprises water and it is extremely important for the survival of all humans and animal species. Water makes up over 326 cubic metres of the planet’s surface, which is almost 71% of its total area out of which 97% is seawater. Only 0.5 percentage of the drinkable water on earth is accessible, while the remaining 2.5 percentage is either trapped in glaciers, polar ice caps, the atmosphere, on soil, is polluted, or lies beneath the earth’s surface far beyond human reach. If the global water supply is 100 L, consequently the amount of drinking water would be only 0.003 L, which is just a teaspoon. Therefore, the management and preservation of drinking water is regarded as a top priority. It is the most critical issue for mankind to address given the extremely limited amount of water that is accessible for use. The quantum of water around the world is represented in Table 1 .

Water is a common and crucial resource shared among all humans, animals, and plants and is a necessity for all species. Each one of these species has its own respective needs for water quality. Total Dissolvable Solids (TDS) of soft water for human consumption range from the best quality stated, which is between 50 mg/dL and 150 mg/dL. Between 150 mg/dL and 300 mg/dL is the next level that can be applied to humans. The plants need water that is between 700mg/dL and 800mg/dL. The animals, especially cattle consume water around the quality of 1000 mg/dL. It is thus evident from all these observations that water quality management is essential to ensure sustainability and a healthy life on Earth. The impact of water quality prediction is crucial at a global level for many reasons. First of all, to get clean and safe water is a basic human necessity and water quality prediction aids to guarantee the availability of potable water for societies worldwide. Water quality is related to public health as polluted water may cause waterborne diseases which could affect millions of humans globally. A sustainable environment is an important aspect of human well-being by preserving ecosystems and biodiversity. The significance of water quality assessment is profound and intricate by various organizations globally. The WHO (World Health Organization) , UNEP (United Nations Environment Programme), EPA (United States Environmental Protection Agency), EEA (European Environment Agency), IWA (International Water Association) and WEF (Water Environment Federation) are fanatical for water quality assessment and addressing the mitigation strategies for water quality challenges. Water quality creates impact on public health globally and resulting in dissemination of waterborne diseases like typhoid, dysentery, cholera, dengue and malaria and cause substantial risks worldwide.

The advancement in computing technologies and artificial intelligence have elevated the standards of water quality assessments 1 . Measurements and estimations about the quality of the water have become easier to calculate and accurate, especially with the development of Industry 4.0 standards and Internet of Things (IoT) sensors. With the integration of IoT sensors, AI solely serves as a supporting tool to automate water quality checks 2 . Classification and Regression models based on machine learning help in determining the water quality. Depending on the outcomes, classification results tend to be binary or multi-classified. Real-time sensor data are collected, given feature labels, and then classified based on the importance of the feature labels. Earlier, these measurements used to be carried out with fuzzy-based decision support systems 3 with subjective decision-making models. AI development has made it possible to classify and analyse quality aspects quantitatively. The accuracy of the water quality assessment has been validated using various performance metrics like accuracy, precision, recall, and f1-score. AI models also support such quantitative analysis, classification of water sources, and prediction of drinkable water as well as identifying the mixing of bouyant pollutants in water sources 4 .

Despite its success in automating tasks and making water quality predictions using diverse models, the AI models lack transparency and are considered black-box where the decisions are derived but the reasoning behind such decisions is not revealed. The present generation validation frameworks for water quality management need justifiability, transparency and explainability, which is possible to be rendered by Explainable AI (XAI) based systems. XAI is a technology that is white-box and answers the uncertainty related to the classification and regression problems of AI. XAI applies a model-agnostic approach, where the machine learning models can be treated independently for interpretation. Additionally, XAI discusses how the model is chosen, how it works, and how it performs categorization. Through the assessment of a problem’s feature weights, XAI also can determine a feature’s relevance. This clarifies how a feature value relates to a certain target class classification. As an example, XAI uses models like Partial Dependency Plots (PDP) 5 , which describes the relationship between the features using lasso functions. This model may identify the linear relationship between two characteristics of water quality data and explain their correlation. In XAI, models like Local Interpretable Model agnostic Explainer(LIME), explain the relationship between a single feature and relevant others in local surrogacy. This infers that, except for the one-row value of the dataset, it is possible to relate a target attribute to the other independent variables. LIME in this regard can be used to explain the target classification for a single row instance about the water quality 6 . In the proposed work, XAI, which employs both local and global surrogates, includes SHAPELY. The model offers a solution that takes into account the importance of each feature in determining the target as well as the dependency between features, the relationship between features, and the explanation of decisions through a variety of plots, including force plots, summary plots, dependency plots, and decision plots. The framework is very adaptable and capable of giving a thorough explanation of the characteristics of the water quality and how they affect the classification of the water quality.

Advantages of the proposed model

Explainable AI plays an important role in improving the interpretability of predictions made by machine learning models. More transparent predictions are generated by these models. In the proposed approach, the authors have employed LIME and SHAP to interpret predictions achieved from machine learning, which identifies inputs as an important metric for selecting the features. By applying the XAI approach, the proposed model provides deep insights into the features and allows informed decision-making in water management processes.

Contributions of the paper

The following points describe the contribution of the proposed work.

The proposed work offers a comprehensive analysis and white-box description of the classification problem for water quality.

The framework incorporates extensive pre-processing of the dataset to ensure it fit to be fed into the XAI model.

Imputation of missing data is carried out to increase the accuracy of the findings.

The proposed work ensures achievement of most significant features, identification of the feature importance, feature dependencies, and feature weights, that enable optimized classification of water quality dataset.

The proposed approach employs both model-based and model-agnostic interpretations, using model-based ML implementations and model-agnostic XAI implementations.

Organization of the paper

Section “ Introduction ” of the paper introduces the problem of the research paper with the description of the unique contributions. Section. Introduction ” also describes the literature review of the related problems on water quality, in related works subsection, with an exhaustive survey of the various applications and case studies pertaining to water quality management using AI and machine learning approaches. Section “ System model and architecture ” describes the methods applied in the proposed work with the implementation of the mathematical model with the algorithm of the proposed work. Section “ Results ” describes the results of various ML and XAI models with relevant tables and graphs. Section “ Discussion ” provides the comparative analysis of the results with a discussion of challenges and solutions of the proposed work. Section “ Conclusion ” concludes the paper with future directions.

Related works

Lu et al. 7 proposed the central environmental protection inspection (CEPI), which was implemented and the causes of transboundary water contamination were investigated. The triple difference technique (DDD) was used to assess how the CEPI affected pollution and the results to determine how significantly water pollution was decreased as well as the significance of CEPI laws for addressing transboundary pollution. Halder et al. 8 , the Turag River’s neighbouring communities are suffering from major health problems as a result of water contamination. For the sustainability of household and aquatic life, the river’s water quality was unsuitable. The study noted that the threshold values for turbidity, total dissolved solids (TDS), chloride (CL-), chemical oxygen demand (COD), carbon dioxide (CO2), and biochemical oxygen demand (BOD) are higher than the standard permissible limits, which may result in health problems like respiratory illnesses, diarrhoea, cholera, dengue, malaria, anaemia, and skin problems. A study evaluating metal pollution management and mitigation tactics on soil and water was presented by Wang et al. 9 . In this study, the remediation of metal contamination from water and soil utilising chemical, physical, and biological approaches was discussed. In this study, the current methods for reducing heavy metal pollution of the soil and water are examined. Elehinafe et al. 10 discussed the importance of water contamination and examined the main cause of water scarcity. The proposed work discussed the effect of hazardous chemicals on the water, including pesticides, heavy metals, and micro-pollutants. This study outlined the numerous technologies that are currently available to eliminate hazardous materials and provide sustainable clean water resources. Mu et al. 11 proposed a solution for the investigation into farmers’ readiness to implement Rural Water Pollution Control (RWPC). This study examines farmers’ viewpoints to improve the quality of life for locals who reside in rural regions and avoid water contamination. To analyse the contributions of contaminants, Wang et al. 12 developed a unique contaminant flux variable model for river water quality assessment. The framework effectively identified the sources of pollution and evaluated the efficacy of projects designed to reduce water pollution. Zadeh et al. 13 proposed WQPs for estimating chemical oxygen demand and biochemical oxygen demand using the MKSVR algorithm. PSO algorithm is used for solving optimization problems. The multiple kernel support vector regression (MKSVR) is compared with SVR and Random Forest Regression and achieves a better accuracy level for BOD prediction. Nagaf et al. 14 presented a framework for assessing the WQI values based on the NSF guidelines. This framework uses four data-driven models such as EPR, M5 MT, GEP and MARS for predicting WQI values in the Karun River. The classification uses 12 water quality parameters and missing values were extracted from the image analysis. Zadeh et al. 15 proposed a model that utilizes gene expression programming, evolutionary polynomial regression, and model trees for predicting WQPs. The biochemical oxygen demand, dissolved oxygen and chemical oxygen demand are used for estimation with nine parameters. The gamma test is used for determining important parameters. Najaf et al. 16 proposed a water quality predicting framework for estimating the water quality index in the Hudson River based on Canadian Council of Ministers of the Environment (CCME) guidelines. The four artificial intelligence techniques M5 MT, Multivariate Adaptive Regression Spline, Evolutionary Polynomial Regression and Gene Expression Programming are used with Landsat 8 OLI-TIRS images. The results proved that the MARS technique achieved the best outcome compared to other models.

Chowdhury et al. 17 emphasized the sources of water contamination which are caused by densely populated industrial areas that are located close to water bodies. The main causes of water contamination are dangerous chemicals and heavy metals. Farmers’ pre-owned pesticides, including different types of carbamate and organophosphorus pesticides, are the main causes of water contamination on agricultural grounds as per the study. Ahivar et al. 18 examined the use of heavy metal pollution indices (HPIs) in soil, water, and sediments. For assessing metal contamination, HPI is considered a crucial instrument. Each method’s pollution index is assessed to interpret the pollution levels. The selection of HPIs based on the parameters and standards for evaluating the quality of the water and soil is offered. Chen et al. 19 presented a study by used various mathematical and statistical approaches to check the quality of water. The factors indicating the water pollution and the seasonal characteristics are evaluated to reduce the river water pollution. The Principal Component Analysis, Cluster Analysis, Network Analysis and Co-Occurrence Analysis were carried out to find the potential source of river water pollution. Fan et al. 20 examined the quality of water using several mathematical and statistical techniques. To lessen river water pollution, the variables implicating contamination and the seasonal traits are assessed. To identify a likely cause of river water pollution, the Principal Component Analysis, Cluster Analysis, Network Analysis, and Co-Occurrence Analysis were performed. Wang et al. 21 formulated the performance indices for explaining the Water-Energy-Pollution nexus (InWEP) effects of scales. The Nexus Pressure Index (NPI) and Nexus Coupling Index (NCI) were used to represent the pollution pressure and the interacted relations. The factors for InWEP were analysed using the Structural Equation Model (SEM) considering four objects namely enterprises, countries, industrial zones and cities. The performance of InWEP was evaluated for the performance metrics - efficiency, structure and location. To evaluate the quality of groundwater surrounding nearby areas in an industrial metropolis, Asomaku 22 evaluated the water pollution indices. Nine samples from three landfills are used in the analysis of the groundwater’s chemical and metal characteristics. The study in Balaram et al. 23 explored many elements that have an impact on water quality, including climate change, industry, aquaculture, mining, and agriculture. For the quantitative and qualitative evaluation of hazardous metals, metal species, isotopes, and other contaminants that are present in water, various ICP-MS techniques are applied. Yuan et al. 24 proposed a water quality monitoring framework using biological sensors for water quality assessment. Borzooei et al. 25 presented a study to estimate the frequency weather events that creates impact on waste water assessment. The Time series data mining approach is used for categorizing the dry and wet weather events. Noori et al. 26 presented a report on decline of groundwater recharge in Iran. The study presents the average amount of ground water recharge is more than the annual runoff 4 utilized WCSPH (A weakly compressible smoothed particle hydrodynamics) model for simulating the near-shore hydrodynamics. The study conducted experimental and numerical evaluation for detecting the causes for mixing the buoyant pollutants in coastal water source. Yeganeh-Bakhtiar 27 presented a framework using MOS (Model Output Statistics) for establishing the statistical relationships among predicator and predicant.

When evaluating water quality using factors like toxicity and pollutants, computer vision and biological sensor systems are utilised in tandem. To retrieve the important data from images taken by a microscope, a microfluidic chip with sensors is utilised. This chip monitors water samples. Figure 1 describes various factors causing water pollution in smart cities including construction activities, atmospheric deposition, natural factors, municipal wastewater, stormwater runoff, incorrect waste disposal, industrial discharges, agricultural runoff, and municipal wastewater. Jeihouni et al. 28 implemented and compared five data mining techniques, including the Ordinary Decision Tree (ODT), Random Forest (RF), Chi-square Automatic Interaction Detector (CHAID), Iterative Dichotomiser 3 (ID3), and Random tree, to identify high-quality water zones. Eight parameters are used in the evaluation process while deriving rules. Compared to the remaining models, the RF performed well, with an accuracy rate of 97.10%. Lee et al. 29 implemented a framework for evaluating the quality of groundwater utilising a Self-Organizing Map (SOM) technique and fuzzy c-means clustering (FCM) was given. The two methods are employed to describe the complex nature of groundwater. SOM employed 91 neurons to categorise 343 groundwater samples, and FCM grouped the water sources into three groups. Agarwal et al. 30 proposed AI based water evaluating technique to predict the water quality index using Particle Swarm Optimization (PSO), Naïve Bayes Classifier (NBC), and Support vector machine (SVM). PSO was used in this regard for optimizing the classifiers wherein the PSO-optimized NBC obtained 92.8% accuracy and PSO-optimized SVM obtained 77.60% accuracy. Table 3 illustrates various existing state-of-art techniques proposed for assessing water quality, its advantages and research gaps.

Figure 1 illustrates the factors causing water pollution. The factors includes Industrial discharges, agricultural runoff, municipal waste water, storm water, improper waste disposal, oil spills and chemical spills, construction wastages, and atmospheric deposition. The factors are very crucial to protect public health and ecosystem , sustainability development, creating public awareness and for pollution prevention.

Factors causing water pollution.

Figure 2 depicts the required physical parameters such as Temperature, Turbidity, Conductivity, Odour and Color represented in percentage, for evaluating the quality of water. Examining the physical parameters is essential for identifying the potential hazards that leads to poor water quality and for preventing ecosystem health.

Physical Parameters.

Figure 3 depicts the necessary chemical parameters, such as pH, Dissolved Oxygen (DO), Total Dissolved Solids (TDS), Nutrients (nitrogen and phosphorus), Total Suspended Solids (TSS), Heavy Metals, and Organic Matter (OM), as well as Chemical Oxygen Demand (COD) and Biochemical Oxygen Demand (BOD) with percentages, that must be measured in order to assess the water’s quality.

Chemical parameters.

Figure 4 presents various supervised learning models for estimating water quality, including Random Forest, Support Vector Machine (SVM), Decision Trees, Neural Networks, and Gradient Boosting Approaches like XGBoost and AdaBoost.

Supervised learning models.

Figure 5 represents various unsupervised learning models such as Principal Component Analysis, Cluster Analysis and Self-Organizing Maps (SOM) for addressing the quality of the water. PCA is a dimensionality reduction approach mainly utilized for analyzing the high dimensional datasets. Cluster analysis techniques are used primarily for grouping water samples based on similarities. SOM technique is principally used for organizing the water quality data.

Unsupervised learning models.

Figure 6 highlights the various Hybrid ML models such as ensemble models with Reinforcement Learning (RL) for addressing the evaluation of quality of water. The various machine learning models can be verified based on the applications, parameters in order to determine the quality of the water, dataset size and its quality based on the assessment of the performance metrics.

Hybrid ML models.

The motivation for the proposed research, along with the research gap analysis with similar existing research works is discussed as per Table 2 . The comparative analysis and research of similar existing works are presented in Table 3 . These two discussions provide a comprehensive understanding of the requirements, that are essentially required in the design of the proposed system and implementation.

Table 3 refers to similar literature review of various models of machine learning such as DT,RF,DCF, SVM, and so on. This table also discusses about various deep learning models such as, Artificial Neural Networks (ANN), Probablistic Neural Network (PNN), Convolution Neural Networks (CNN) and statistical regression models such as Auto-Regression in Moving Average(ARIMA). This table discusses the the research gaps identified and enhanced in the proposed work. These models were mostly numerical evaluations with regression analysis. The proposed model and the system is classifier which deploys XAI framework, to discuss the impact of parameters, that determine the portability of the water with end user perspective. This is towards achieving environmental sustainability on water conservation and harvesting.

Statement of objectives

The proposed work offers a comprehensive analysis and white-box description of the classification problem for water quality . The framework incorporates extensive pre-processing of the dataset to ensure it fits into the XAI model. Imputation of missing data is carried out to increase the accuracy of the findings. The proposed work ensures the achievement of the most significant features, identification of the feature importance, feature dependencies, and feature weights, that enable optimized classification of the water quality dataset. The proposed approach employs both model-based and model-agnostic interpretations, using model-based ML. Donnelly et al. 46 implementations and model-agnostic XAI implementations. The quality of water is greatly challenged by innumerable influencing factors. These factors vary from condition to condition and place to place. For example, Microplastics (MP) are emerging pollutants in the marine environment with potential toxic effects on littoral and coastal ecosystems 47 and as well as identifying the mixing of bouyant pollutants in water sources 4 . The laboratory evaluations show the presence of polyethene (PE) particles in the waves of the ocean with wave steepness Sop of 2–5%. The transportation of which could cause severe water pollution on the seashores 48 .These measurements require quantification and feature analysis when it is evaluated with AI. This is where the XAI plays a vital role in measuring the order and degree of the pollutants causing the quantifiable pollution in the water.

Case studies

Importance of XAI in Water Quality Assessment: The following case studies delineate the advent of the potential impact of XAI, with a groundbreaking revolution in water quality assessment.

Case Study 1: Pollution of Ganges 49 This case study emphasises the Ganga River pollution issue in India, which has an extremely detrimental impact on humans and the entire ecosystem. The Ganga River is polluted by industrial, animal, and human waste. The main source of pollutants is industrial rubber waste, followed by leather and plastic manufacturers who dump their untreated wastewater into the river. The Ganga Action Plan was developed by the Indian government to combat Ganga pollution. This implies the need for the reinforcement of environmental restrictions to improve river quality.

Materials and methods

An effective policy for health protection should thus emphasize providing access to safe drinking water regardless of social and economic diversity. In some places, it is evident from previous studies that investments in access to clean water and sanitation yield economic benefits for any country. It is a significant aspect of eco-friendly health and public safety, as it regulates the appropriateness of water for numerous purposes, such as drinking, agriculture, industry, and recreational purposes. The important key indicators related to water quality are its physical, chemical, and biological characteristics and its sources of pollution. The dependent target class is potability. The other independent features are pH value, hardness, solids (Total Dissolved Solids-TDS), Chloramines, sulfate, conductivity, organic carbon, trihalomethanes, and turbidity. Water’s potability indicates its purity and safety for ingestion. The parameters used and their WHO limits, the hyper-parametric analysis are listed in Table 4 , and the feature description of parameters are listed in Table 5 .

XAI framework facilitates transparent and interpretable explanations of the outcome generated by the ML algorithm-based frameworks. XAI can thus be applied in the present context of water quality assessment to ensure accurate decision-making, thereby, enabling trustworthiness, enhancement of transparency and interpretability of the behaviour of the model.

Hydro-climatic application

XAI framework can be used to solve Hydro-Climatic problems 50 with diverse spatio-temporal scales. XAI is utilized to unveil the nonlinear correlative causes, in which the performance of the model is enhanced. It enables the users to discover new knowledge and further easily understand the rationale behind the decision outcomes.

Groundwater potential predictions

XAI approach can explain the decisions made by ML models for groundwater potential prediction. The user can easily interpret the outcomes and further comprehend the underlying for an outcome in the realm of water quality evaluation for conservation, and sustainability of water management.

Water quality predictions

XAI framework can forecast water quality using metrics and factors with interpretable results. Water quality assessment managers can comprehend the variables and parameters used for outcomes. This forces quality managers to mitigate water quality issues.

Flood hazard risk predictions

Floods can trigger landslides from excessive rainfall. Flooding causes countless casualties and property damage. Disaster warning systems need a flood risk assessment. XAI can forecast rapid water depths and provide timely, interpretable alerts to protect public health and safety.

Environmental impact assessment

XAI approach can be used for assessing the environmental impact on the water pollution incidents, and provide insight for mitigation and management. It enhances transparency and accountability by providing insights into the factors and parameters influencing environmental conditions. The analysis provided by the XAI model helps the stakeholders to identify the most significant factors contributing towards the environmental outcome.

System model and architecture

System model.

Worldwide, numerous water bodies are contaminated by a variety of anthropogenic and natural processes, resulting in a variety of health problems for human life. Thus water quality requires rigorous monitoring and management to prevent pollution. In accordance with WHO guidelines, the polluted water must be treated using the proper water treatment techniques before consumption. The quality of water is contaminated by the incessant addition of toxic chemicals and microbes and also by the relentless addition of local and industrial sewage sludge, trash, and extra hazardous waste that are toxic to humans and society. Many uncertainties are required to be quantified for all machine learning models. The uncertainties such as selecting and gathering the training data, absolute and accurate training data, understanding the machine learning models with performance bounds and drawbacks and finally the uncertainties which are based on the operational data. To minimize the challenges, adhoc steps like studying the model variability and sensitivity analysis are applied. In current years, the validation of water quality has taken active momentum because of ever-increasing water pollutants which spoil water that is dedicated for domestic use and irrigation. Water quality indices (WQIs) are used worldwide very efficiently for the assessment of the quality of both groundwater and other relevant water sources. Machine Learning techniques play a substantial role in identifying the quality of water using explainable AI. Figure 7 depicts the overall architecture of the proposed framework of our study. The dataset used in the study is split into the ratio of 70:30 wherein 70% is used for training and 30% is used for testing. The model is trained using a decision tree, random forest, SVM, logistic regression, and Naive Bayes algorithms. XAI model is implemented in the framework wherein LIME and Shapely are used to provide explainability and interpretability to the results generated by the machine learning model .

Interfacing ML algorithms with XAI.

Decision tree

The decision tree is stated as a recursive partition of the set of all possible instances 27 51 . The goal of a decision tree is to split the data which consequences in maximum information gain 52 . Let L be a sample for learning, L= ( \(v_{1}\) , \(c_{1}\) ), ( \(v_{2}\) , \(c_{2}\) ),( \(v_{i}\) , \(c_{j}\) ). Here, \(v_{1}\) , \(v_{2}\) , \(v_{3}\) , \(v_{i}\) are represented for measurement vectors, and \(c_{1}\) , \(c_{2}\) , \(c_{3}\) , \(c_{j}\) are represented for class labels.The batch conditions are reliant on one of the vector variables denoted as \(s_{i}\) 53 . Let us assume if the \(e_{i}\) of an element fits class label \(c_{i}\) , then \(p_{i}\) is denoted as per the Eq. ( 1 ).

Entropy evaluates the random value from the given samples and the homogeneity of the expected rate of a group of data 54 . To divide the data most optimally, the lowest value of entropy signifies better homogeneity.

L represents the data set evaluated by the entropy, ‘i’ denotes the classes in the set L, and \(e_{i}\) indicates the number of data labels that fit class ’i’ 55 . The least value of entropy is used for choosing the best feature. Information gain enumerates the amount of information provided by a particular characteristic about the target variable to minimize the uncertainty present in the data set. It is calculated by comparing the weighted average of entropy to the original data set after the splitting process. Let us assume that R is the rate for the features ‘f’, \([|{L}^R|]\) denotes the subset of LS so that bf=R 56 . After splitting L on the feature, information gain is given as follows.

The Gini index evaluates the heterogeneity of a selected node in the decision tree. It counts the probability of wrongly identifying data in the node. The Gini index begins from the value 0 to 1, where 0 indicates a pure node and 1 denotes a node that is distributed equally. The Gini index is represented as

Here, \(e_{i}\) represents the quantity of data labels. When the data is divided on class d as L1 and L2 with sizes \(s_{1}\) and \(s_{2}\) , Gini is evaluated as

Due to its comprehensible nature, decision trees can manage both numerical and categorical data with automatic feature selection.

Random forest

Random forest is an ensemble method that groups the results of multiple decision trees to compute predictions with enhanced accuracy. Every decision tree is improved on a random subset of labels from the dataset, to achieve diversity between the trees. When the data in the training label is t, then with replacement ‘n’ data are verified as bootstrap data 57 . This is done to produce the decision tree with training data. When there are ’m’ labels, a \(<<\) m is selected so that ‘a’ values are considered at random from ‘m’. The value ‘a’ is constant when the tree is growing to the highest level. The highest vote is noted as a new instance. (GE*) is the generalization error for the random forest and is denoted as

Here, f(X, Y) is a margin function to count the average number of votes from (X, Y). X denotes the prediction value and Y denotes the classification problem. The margin function is represented as

where ’F’ is for the indicator function. The value for the margin function is indicated as

The average value of a random forest and the mean correlation of the classifiers are combined as generalization errors. The p denotes the mean of the correlation. The generalization error for the upper bound is

Random forest reduces the over-fitting problem compared to a single decision tree. It can effectively manage high-dimensional data.

Support vector machine (SVM)

Let us consider a binary classification problem 1 or −1 to represent the sample variables 58 . When i elements of the sample variable is − 1, it is a positive class. When the i variables of the samples is 1, it is a negative set. Let V_i  = X1, X2,...Xn, Yi, i = 1,2,...n, \(Y\_{i}\in {-1,1}\) , Si indicates i item from the samples. Yi is the i item of the tests performed 59 . To split the samples into two parts, the function f(X) = ZTX+ b is used, where Z is the coefficient vector to normalize the hyperplane. The optimal margin is given as

\(\underbrace{MIN}_{\begin{array}{c} w, b, \\ \varepsilon \end{array}} \left( {\frac{1}{2}}Z^{TZ}+C\sum _{i=1}^{n}\varepsilon _i\right)\)

subject to:

The Lagrangian equation is given as

\(\underbrace{MAX}_{\propto } \left( \sum _{i=1}^{n}{\propto _i-\frac{1}{2}}\sum _{i,j=1}^{n}{\propto _i\propto _jY_iY_jX_iX_j}\right)\)

The Lagrangian equation with the maximum value with \(\propto _i\) a positive multiplier for the equation \(\sum _{i=1}^{n}{\propto _iY_i=0}\) and \(\propto _i\ge 0\) to change the optimal hyperplane 60 is presented. The optimal equation is given as

In the above equation \(\propto _i=0\) of the Lagrangian multiplier is nearest to the margin of the optimal hyperplane denoted as a support vector. This data is linearly separable by the kernel to evaluate the expected result from the instance 61 . The kernel function is denoted as

The generalized linear equation is changed to represent the non-linear dual Lagrangian \(La(\alpha )\) .

\(Lag\left( \propto \right) =\ \sum _{i=1}^{n}{\propto _i-\frac{1}{2}\sum _{i,j=1}^{n}{\propto _i\propto _jY_iY_jK\left( X_i,X_j\right) }}\)

Subject to:

The Lagrangian equation can be used for the separable case as

The SVM algorithm is very effective when the quantity of features is higher than the number of samples 62 .

Logistic regression

Logistic regression is used for binary classification problems to forecast the probability of an occurrence matching to a particular class. If the dependent value is binary, a regression analysis is used. The idea in logistic regression(logreg) is the logarithm ‘logn’ of odds of X, and odds are the ratios of probabilities ‘pb’ of X 63 . The rate of the independent value is termed odds because logistic regression measures the probability of an act that happens over the likelihood of an occurrence that does not happen.

where p is the probability of a positive output and x is the variable. The \(\alpha\) and \(\beta\) , are the logistic regression parameters 64 . The above equation is used for finding the number of occurrences as

\(p=probability(Y=positive\ outcome|X=x,\) a specific value)

For multiple predictors, a logic regression equation can be written as

\(p=probability(Y=positive\ outcome|X_1=x_1,\ldots ,x_k)\)

Here, pb refers to the probability of the positive occurrence of the event, the Y-intercept is \(\alpha\) , the regression coefficient is \(\beta\) , and e is 2.71828. Logistic regression is applied in various domains like finance, healthcare, social sciences, and many more for predicting diseases, credit default, etc.

Naive Bayesian classification

Gaussian Naive Bayes is a probabilistic classification algorithm developed based on Bayes theorem. It refers to the features which represent a normal distribution 65 . It classifies the samples as most likely classified as

If the sample \(Y_{j}\) is a vector, \(x_{j}\) is the \(j^{th}\) value which contains different values of \(y_{j}\) . The attributes used are dependent and it is shown as

Substituting the above equation into Bayes classification, we get

The Gaussian Naive Bayes algorithm is mainly applied for spam filtering, sentiment analysis, and text classification problems where the features must be continuous and follow the Gaussian distribution 66 .

LIME (Local interpretable model-agnostic explanations)

LIME explains the predictions of any kind of classifier by approximating locally along with an interpretable system. It changes the data sample by altering the values of features and monitors the impact of the result. It explains the predictions from every sample 67 . To receive the labels for the current data, alter the samples z ’s into the unique form \(z \in {\mathbb {R}}^d\) . Since the samples x ’ are generated randomly, x samples closer to the unique instance z for weighing are considered. The weight is evaluated as \(\Pi _z(x)\) for measuring the intimacy between the data z to x. The currently weighted data X and the samples formed by f ( x ), are trained as \(g \in G\) , where G is a model. The interpretable model \(\xi (x)\) of the current data g for explaining f ( x ) as

L is the loss function to measure whether g is following the state of f in the nearest neighborhood of z . If the loss function is reduced, the behaviour of g takes the behaviour of f as \(\Pi _z\) . The complexity of the model \(\Omega (g)\) should be low. When \(g(x')\) is considered as a linear function, \(g(x') = \varphi ^T x' + \varphi _0\) , changes the equation into a linear regression task to evaluate \(\varphi\) and \(\varphi _0\) .

SHAP (SHAPELY Additive exPlanations)

SHAP values determine the status of each feature for the prediction of a specific class 68 . The prediction f ( y ), using \(s(y')\) , a model for the binary elements \(x' \in \{0,1\}^M\) with the sets \(\emptyset _i \in {\mathbb {R}}\) , is given as

M refers to the explanation variable.

where f is the model of the SHAP, z refers to the variable, and \(z'\) are the variables chosen. The value \(f_y(x') - f_y(x'\setminus i)\) indicates all the predictions.

In this section two algorithms are discussed: one for the algorithm-based evaluation of water quality 1 and another for the algorithm-based explanation of water quality 2 . These two algorithms provide a holistic analysis and explanation of water quality management.

Algorithm for water quality classification

Algorithm for water quality Explanation

The water quality is assessed in the proposed work based on nine parameters such as pH value, Hardness (Total Dissolved Soils), Sulphate, Chloramines, Trihalomethanes, Conductivity, Organic carbon, and Turbidity. The target class for this dataset is Potability which is binary where 0 indicates that the water is not potable and 1 reflects its potability.

The dataset consisted of high missing values on sulphate and lower missing values on Chloramines and Trihalomethanes. The missing value imputation is hence performed and all the attributes are imputed for the missing values. The target class is converted into a numeric array for the processing of XAI models. This is done with the label encoder application of Python. The dataset is split with a ratio of 80:20 for training and testing.

The correlation analysis is performed on the dataset. The attribute Hardness has a high correlation of 0.34 with the target attribute potability. The next best correlation value is 0.24, which is rendered by the attribute Chloramines, followed by 0.21 produced by the Trihalomethanes attribute. Turbidity is the next better parameter with a correlation value of 0.16. The correlation heat map between the attributes of interest and the target attribute is presented in Fig. 8 .

Correlation analysis for water quality attributes.

The trained dataset is applied with SVM, LR, DT, RF and Gaussian Naive Bayes machine learning models. The SVM did not provide the desired classification and failed to converge for the portable data. The other models generated the results within the desired range and are presented in Table 6 .

The sensitivity and specificity measurements for the Machine learning models are presented in Table 7 . Considering the performance metrics, the results reveal the superiority of the RF model which generates a better outcome in comparison to the other models and thus it has been selected to be fed into the XAI model to provide enhanced interpretability, justifiability and transparency.

The XAI model implementation is performed considering SHAPELY values in the pandas’ application. This application focuses on the value of each feature in determining the target attribute which is potability. The significance of every feature is assessed through the various applications of SHAPELY. The first XAI model generated is the force plot, which provides the minimum and maximum prediction score of the target attribute in a dataset. The blue colored contour shows that a low score is measured and the red color shows a high score. The values at the separation boundary have the highest priority attribute. The force plot is presented in the Figs. 9 and 10 .

Force plot for water quality.

Force plot for potability.

The Global surrogate version of the force plot is presented in Fig. 11 . The blue regions indicate no potability and the red-coloured regions indicate potability. The border areas of the intersection show the attributes which have higher significance for the feature selection. The Sulphate value of 444 at the point of intersection indicates its significance in explaining this test patch for the entire dataset.

Test patch for potability.

The next XAI application of SHAPELY is the summary plot. This plot describes the features in determining binary classification problems. This predicts the scale of low to high for two significant results. The blue contour indicates lower significance towards the prediction and red indicates higher significance. The summary plot is shown in Fig. 12 . The Solids, pH, Sulfate, and Hardness show higher significance in determining the output.

Summary plot for potability.

The dependency plot shows the relationship between two features in the dataset. It provides the output in granular form with a variable-like result rather than simply a graph-like result of a Partial Dependency Plot(PDP). The relationship between the Sulphate and Potability is depicted in Fig. 13 . The mid-range of the dataset provides more granular output, which shows that the Sulphate parameter values are more significant in determining the values of potability in the mid-range of the dataset.

Dependency plot for potability.

The decision plot, which displays how the values of the features affect the goal, is the final model of XAI. This plot is a local surrogate plot, which would only explain a certain data instance, in which what values of the attributes influence the decision to be 1 or 0 as the decision of the model. The decision plot for the potability as 1 is illustrated in Fig. 14 . The potability 0 is illustrated in Fig. 15 .

Decision plot for potability.

The results of the experiment reveal the superiority of the RF model which generates an accuracy of 0.999 followed by DT, generating an accuracy of 0.998. The lowest accuracy is generated by the SVM model of 0.63. The RF is thus chosen for the implementation of the XAI model using SHAPELY. The comparative analysis of the aforementioned various models is depicted in Fig. 16 , considering evaluation metrics accuracy, precision, recall, and f1-score. In the case of all the performance metrics, the RF model outperforms the other models. Figure 17 shows the comparison of the sensitivity and specificity measures. The RF model stands superior in these considerations as well. Thus, the discussion offers a visual representation and justification of the reasoning behind the choice of RF to be included in the XAI framework to offer explainability.

Comparative analysis of machine learning models used.

Comparative analysis of sensitivity and specificity.

Apart from the selection of the RF model, SHAPELY provided five different models to explain the feature importance and relationships. The proposed work presented the force plot, summary plot, test patch, dependency plot, and decision plot. The Final decision plot explained how the classification is carried out using the corresponding values of the independent variables. Thus the black-box classification is explained in the white-box context of XAI. The following section describes the challenges and opportunities of the proposed work with an emphasis on future directions.

The proposed work may be influenced by the following challenges which are described in detail as follows,

Global unity

For the successful implementation of the system, a unanimously accepted implementation is essential. Unfortunately, water quality estimation and related research are limited to consideration of specific datasets acquired for a particular region, wherein the generated results may differ with the changes in geographic location. Thus the generated results can never be considered suitable on a global scale. The parameters that influence the water quality may also vary across the world, and hence the proposed work can never be considered as a universal solution.

Training and re-training

The qualifying attributes that determine the quality of water vary across the globe and hence the proposed model needs to be re-trained 69 when applied to a new environment of study. This would allow the model to unlearn and re-learn new environments. On the contrary, the complexity of the model would also increase. The accuracy and other performance metrics which are measured in the proposed work may drastically decrease as well in a different environment of study. Thus applying this model to versatile environments is complex and would be a challenging task.

Subjective or quantitative

The trade-off from subjective analysis (which was done through fuzzy-based methods in the form of the Analytical Hierarchy Process (AHP) and The Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS)) has improved the performance and ability to classify the models with better accuracy. However, the involvement of a subject matter expert is a missing point in the current research. Despite all the implementation and analysis from an engineering perspective, the involvement of an environmental scientist in any aspect of water research would contribute towards the enhancement of research quality.

Confusing solids

The proposed work identifies Solids as the primary influencing factor that affects potability. In real-world applications, solids can be of any form. For example, in sewage water treatment plants it can be either mud, Fat-Oil-Grease(FOG), or any other substances. Every solid wastage has its way of filtration and impact on water quality, which makes the recordings unstable from time to time. The attributes of research are too complex to handle in real-life scenarios, which acts as an inevitable yet detrimental impact.

Environmental challenges

Water resources are under serious threat due to water scarcity, water contamination, water conflicts and climate changes. Chemical and the municipal wastewater contaminates the water and endangering the life of the aquatic organisms and affect their ability to reproduce. This also makes them an easier prey to their predators. The food cycle and livelihood of the human is also greatly affected by the water contamination. Chemical substances make the water hard to recycle and consume by reducing the regeneration ratios.

Water quality and industrial sustainability

The era of Industry 5.0 focuses on the consumer centric industrial evolution with the idea of environmental sustainability. The futuristic technologies evolve with the improvement of technical viability, with the mission of sustainable development in the environmental aspects. Since the water is an irreplaceable and finite, the demand of the water is increasing with the industrial evolution and the water requirements on manufacturing and production industries would be very much essential as ever. The challenge is enhancement of the water harvesting, recycling and conservation. For all the above said processes quality of the water is the common essential requirement. Thus the quality of the water is more critical in all futuristic technological developments.

Research finding of the proposed work

The following items are presented as the findings are outcomes of the proposed work

The proposed work performs an exploratory analysis with XAI implementation providing an ability to improve the reliability of machine learning models providing explanation and transparency to the classification process.

The proposed work acquires data from a single dataset, where the performance of classification yields optimized results. This result may vary if the model is subjected to a different dataset constituting different features and instances.

The XAI reveals the most significant features contributing towards classification results and also explains the same.

The best fitting machine learning model is chosen for the explanation through an exhaustive analysis and evaluation of all the models considering the essential performance metrics. Thus the results produced by SHAPELY can be considered as the most reliable and acceptable. 

The proposed work also suggests the importance of the subject matter expert, which can extend the usability of the proposed model at the universal level.

The predictions of the proposed work with the support of an explainer, helps end users and consumers to understand the quality of the water they use.

The features related to the classification and explanation, can be further controlled to diminish the levels of chemicals and pollutants in water recycling.

Total dissolvable solids quantification and the feature weights for the same determine the levels of filtration and carbon purification required in the recycling plants.

The proposed work brings insights of pollutants on the seashore and how the explainabilty can support the impurity estimations for such conditions also.

Water quality management impacts almost all aspects of life on earth and clean water is a basic necessity. The proposed work is extremely relevant in this regard wherein an exploratory analysis conducted to analyze and control the factors that deteriorate the quality of the water. The impact of these factors is explained using XAI models. The contribution of the XAI model lies in its ability to explain the role of the underlying parameters towards the classification of water being potable or not, based on their relative importance and unique properties. The XAI model uses SHAPELY considering the probabilistic prediction generated from the Random Forest classifier. This RF model in this regard is chosen as it yields the highest accuracy of 0.999 with sensitivity and specificity of 0.999 and 0.998, which is found to be superior in comparison to the other state-of-the-art models considered in the study. This justifies the reason for the RF to be selected for XAI implementation. The proposed model identifies the parameter “solid” as the most significant in terms of its impact on the potability of water. The proposed model yields optimized and explainable results considering the dataset used in the study. Future work may involve more complex and heterogeneous datasets to generate predictions. In such scenarios, the metric evaluations may differ. The usage of deep learning algorithms could further enhance the examination the solid sediments and generate classification results based on their mass, dimensions, and shape. The use of XAI in such a model would ensure a better explanation of factors relevant to the solid sedimentation in water.

Data availability

The data that support the findings of this study are available from the corresponding author, upon reasonable request.

Zhu, M. et al. A review of the application of machine learning in water quality evaluation. Eco-Environ. Health 1 , 107–116. https://doi.org/10.1016/j.eehl.2022.06.001 (2022).

Article   PubMed   PubMed Central   Google Scholar  

Miller, M., Kisiel, A., Cembrowska-Lech, D., Durlik, I. & Miller, T. Iot in water quality monitoring are we really here?. Sensors 23 , 960. https://doi.org/10.3390/s23020960 (2023).

Article   ADS   PubMed   PubMed Central   Google Scholar  

Akhtar, N. et al. Modification of the water quality index (wqi) process for simple calculation using the multi-criteria decision-making (mcdm) method: A review. Water 13 , 905. https://doi.org/10.3390/w13070905 (2021).

Article   CAS   Google Scholar  

Abolfathi, S. & Pearson, J. Application of smoothed particle hydrodynamics (sph) in nearshore mixing: A comparison to laboratory data. Coastal Eng. Proc. 35 , 1–13 (2017).

Google Scholar  

Hájek, M. et al. A European map of groundwater ph and calcium. Earth Syst. Sci. Data 13 , 1089–1105. https://doi.org/10.5194/essd-13-1089-2021 (2021).

Article   ADS   Google Scholar  

Li, L. et al. Interpretable tree-based ensemble model for predicting beach water quality. Water Res. 211 , 118078. https://doi.org/10.1016/j.watres.2022.118078 (2022).

Article   CAS   PubMed   Google Scholar  

Lu, J. Can the central environmental protection inspection reduce transboundary pollution? Evidence from river water quality data in china. J. Clean. Prod. 332 , 130030 (2022).

Halder, J. N. & Islam, M. N. Water pollution and its impact on the human health. J. Environ. Hum. 2 , 36–46 (2015).

Article   Google Scholar  

Wang, Z. et al. Overview assessment of risk evaluation and treatment technologies for heavy metal pollution of water and soil. J. Clean. Prod. 379 , 134043 (2022).

Elehinafe, F. B., Agboola, O., Vershima, A. D. & Bamigboye, G. O. Insights on the advanced separation processes in water pollution analyses and wastewater treatment: A review. S. Afr. J. Chem. Eng. 48 , 188–200 (2022).

Mu, L., Mou, M., Tang, H. & Gao, S. Exploring preference and willingness for rural water pollution control: A choice experiment approach incorporating extended theory of planned behaviour. J. Environ. Manag. 332 , 117408 (2023).

Wang, Y., Ding, X., Chen, Y., Zeng, W. & Zhao, Y. Pollution source identification and abatement for water quality sections in Huangshui River Basin, China. J. Environ. Manag. 344 , 118326 (2023).

Najafzadeh, M. & Niazmardi, S. A novel multiple-kernel support vector regression algorithm for estimation of water quality parameters. Nat. Resour. Res. 30 , 3761–3775 (2021).

Najafzadeh, M., Homaei, F. & Farhadi, H. Reliability assessment of water quality index based on guidelines of national sanitation foundation in natural streams: Integration of remote sensing and data-driven models. Artif. Intell. Rev. 54 , 4619–4651 (2021).

Najafzadeh, M., Ghaemi, A. & Emamgholizadeh, S. Prediction of water quality parameters using evolutionary computing-based formulations. Int. J. Environ. Sci. Technol. 16 , 6377–6396 (2019).

Najafzadeh, M. & Basirian, S. Evaluation of river water quality index using remote sensing and artificial intelligence models. Remote Sens. 15 , 2359 (2023).

Chowdhury, M. A. Z. et al. Organophosphorus and carbamate pesticide residues detected in water samples collected from paddy and vegetable fields of the Savar and Dhamrai Upazilas in Bangladesh. Int. J. Environ. Res. Public Health 9 , 3318–3329 (2012).

Article   CAS   PubMed   PubMed Central   Google Scholar  

Ahirvar, B. P., Das, P., Srivastava, V. & Kumar, M. Perspectives of heavy metal pollution indices for soil, sediment, and water pollution evaluation: An insight. Total Environ. Res. Themes 6 , 100039 (2023).

Chen, K., Liu, Q.-M., Peng, W.-H., Liu, Y. & Wang, Z.-T. Source apportionment of river water pollution in a typical agricultural city of Anhui province, Eastern China using multivariate statistical techniques with apcs-mlr. Water Sci. Eng. 16 , 165–174 (2023).

Fan, S. et al. Improved multi-criteria decision making method integrating machine learning for patent competitive potential evaluation: A case study in water pollution abatement technology. J. Clean. Prod. 403 , 136896 (2023).

Wang, Z., Wang, C. & Liu, Y. Evaluation for the nexus of industrial water-energy-pollution: Performance indexes, scale effect, and policy implications. Environ. Sci. Policy 144 , 88–98 (2023).

Asomaku, S. O. Quality assessment of groundwater sourced from nearby abandoned landfills from industrial city in Nigeria: Water pollution indices approach. HydroResearch 6 , 130–137 (2023).

Balaram, V., Copia, L., Kumar, U. S., Miller, J. & Chidambaram, S. Pollution of water resources and application of icp-ms techniques for monitoring and management: A comprehensive review. Geosyst. Geoenviron. 2 , 100210 (2023).

Yuan, F., Huang, Y., Chen, X. & Cheng, E. A biological sensor system using computer vision for water quality monitoring. Ieee Access 6 , 61535–61546 (2018).

Borzooei, S. et al. Impact evaluation of wet-weather events on influent flow and loadings of a water resource recovery facility. In New Trends in Urban Drainage Modelling: UDM 2018 11 706–711 (Springer, 2019).

Noori, R. et al. Decline in Iran’s groundwater recharge. Nat. Commun. 14 , 6674 (2023).

Article   ADS   CAS   PubMed   PubMed Central   Google Scholar  

Yeganeh-Bakhtiary, A., EyvazOghli, H., Shabakhty, N., Kamranzad, B. & Abolfathi, S. Machine learning as a downscaling approach for prediction of wind characteristics under future climate change scenarios. Complexity 2022 , 8451812 (2022).

Jeihouni, M., Toomanian, A. & Mansourian, A. Decision tree-based data mining and rule induction for identifying high quality groundwater zones to water supply management: a novel hybrid use of data mining and gis. Water Resour. Manag. 34 , 139–154 (2020).

Lee, K.-J. et al. The combined use of self-organizing map technique and fuzzy c-means clustering to evaluate urban groundwater quality in Seoul Metropolitan City, South Korea. J. Hydrol. 569 , 685–697 (2019).

Article   ADS   CAS   Google Scholar  

Agrawal, P. et al. Exploring artificial intelligence techniques for groundwater quality assessment. Water 13 , 1172 (2021).

Wang, Y. et al. Monthly water quality forecasting and uncertainty assessment via bootstrapped wavelet neural networks under missing data for Harbin, China. Environ. Sci. Pollut. Res. 20 , 8909–8923 (2013).

El Bilali, A., Taleb, A. & Brouziyne, Y. Groundwater quality forecasting using machine learning algorithms for irrigation purposes. Agric. Water Manag. 245 , 106625 (2021).

Arabgol, R., Sartaj, M. & Asghari, K. Predicting nitrate concentration and its spatial distribution in groundwater resources using support vector machines (svms) model. Environ. Model. Assess. 21 , 71–82 (2016).

Sajedi-Hosseini, F. et al. A novel machine learning-based approach for the risk assessment of nitrate groundwater contamination. Sci. Total Environ. 644 , 954–962 (2018).

Article   ADS   CAS   PubMed   Google Scholar  

Ransom, K. M., Nolan, B. T., Stackelberg, P., Belitz, K. & Fram, M. S. Machine learning predictions of nitrate in groundwater used for drinking supply in the conterminous united states. Sci. Total Environ. 807 , 151065 (2022).

Yadav, B., Gupta, P. K., Patidar, N. & Himanshu, S. K. Ensemble modelling framework for groundwater level prediction in urban areas of India. Sci. Total Environ. 712 , 135539 (2020).

Tomić, A. Š, Antanasijević, D., Ristić, M., Perić-Grujić, A. & Pocajt, V. A linear and non-linear polynomial neural network modeling of dissolved oxygen content in surface water: Inter-and extrapolation performance with inputs’ significance analysis. Sci. Total Environ. 610 , 1038–1046 (2018).

Zhi, W. et al. From hydrometeorology to river water quality: Can a deep learning model predict dissolved oxygen at the continental scale?. Environ. Sci. Technol. 55 , 2357–2368 (2021).

Srinivas, R., Bhakar, P. & Singh, A. P. Groundwater quality assessment in some selected area of Rajasthan, India using fuzzy multi-criteria decision making tool. Aquat. Procedia 4 , 1023–1030 (2015).

Haghibi, A. H., Nasrolahi, A. H. & Parsaie, A. Water quality prediction using machine learning. J. Water Qual. Res. 53 , 3–13 (2018).

Liu, M. & Lu, J. Support vector machine-an alternative to artificial neuron network for water quality forecasting in an agricultural nonpoint source polluted river?. Environ. Sci. Pollut. Res. 21 , 11036–11053 (2014).

Chen, K. et al. Comparative analysis of surface water quality prediction performance and identification of key water parameters using different machine learning models based on big data. Water Res. 171 , 115454 (2020).

Sagan, V. 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. 205 , 103187 (2020).

Wu, Y., Zhang, X., Xiao, Y. & Feng, J. Attention neural network for water image classification under iot environment. Appl. Sci. 10 , 909 (2020).

Pu, F., Ding, C., Chao, Z., Yu, Y. & Xu, X. Water-quality classification of inland lakes using landsat8 images by convolutional neural networks. Remote Sens. 11 , 1674 (2019).

Donnelly, J., Daneshkhah, A. & Abolfathi, S. Forecasting global climate drivers using gaussian processes and convolutional autoencoders. Eng. Appl. Artif. Intell. 128 , 107536 (2024).

Abolfathi, S., Cook, S., Yeganeh-Bakhtiary, A., Borzooei, S. & Pearson, J. Microplastics transport and mixing mechanisms in the nearshore region. Coast. Eng. Proc. https://doi.org/10.9753/icce.v36v.papers.63 (2021).

Stride, B., Abolfathi, S., Odara, M. G. N., Bending, G. D. & Pearson, J. Modeling microplastic and solute transport in vegetated flows. Water Resour. Res. 59 , e2023WR034653. https://doi.org/10.1029/2023WR034653 (2023).

Unacademy (2022).

Başağaoğlu, H. et al. A review on interpretable and explainable artificial intelligence in hydroclimatic applications. Water 14 , 1230 (2022).

Habib, M., O’Sullivan, J., Abolfathi, S. & Salauddin, M. Enhanced wave overtopping simulation at vertical breakwaters using machine learning algorithms. PLoS ONE 18 , e0289318 (2023).

Mpia, H., Mburu, L. & Mwendia, S. Applying data mining in graduates’ employability: A systematic literature review. Int. J. Eng. Pedag. 13 , 86–108. https://doi.org/10.3991/ijep.v13i2.33643 (2023).

Raileanu, L. E. & Stoffel, K. Theoretical comparison between the gini index and information gain criteria. Ann. Math. Artif. Intell. 41 , 77–93. https://doi.org/10.1023/b:amai.0000018580.96245.c6 (2004).

Article   MathSciNet   Google Scholar  

Gulati, P., Sharma, A. & Gupta, M. Theoretical study of decision tree algorithms to identify pivotal factors for performance improvement: A review. Int. J. Comput. Appl. 141 , 19–25. https://doi.org/10.5120/ijca2016909926 (2016).

Tangirala, S. Evaluating the impact of GINI index and information gain on classification using decision tree classifier algorithm. Int. J. Adv. Comput. Sci. Appl. 11 , 110277. https://doi.org/10.14569/ijacsa.2020.0110277 (2020).

Xu, P. Review on studies of machine learning algorithms. J. Phys. 1187 , 052103. https://doi.org/10.1088/1742-6596/1187/5/052103 (2019).

Purwanto, A. D., Wikantika, K., Deliar, A. & Darmawan, S. Decision tree and random forest classification algorithms for mangrove forest mapping in Sembilang National Park, Indonesia. Remote Sens. 15 , 16. https://doi.org/10.3390/rs15010016 (2022).

Huang, H. et al. A new fruit fly optimization algorithm enhanced support vector machine for diagnosis of breast cancer based on high-level features. BMC Bioinform. https://doi.org/10.1186/s12859-019-2771-z (2019).

Ji, Y. & Sun, S. Multitask multiclass support vector machines: Model and experiments. Pattern Recogn. 46 , 914–924. https://doi.org/10.1016/j.patcog.2012.08.010 (2013).

Übeyli, E. D. ECG beats classification using multiclass support vector machines with error correcting output codes. Dig. Signal Process. 17 , 675–684. https://doi.org/10.1016/j.dsp.2006.11.009 (2007).

Cortes, C. & Vapnik, V. Support-vector networks. Mach. Learn. 20 , 273–297. https://doi.org/10.1007/bf00994018 (1995).

Ye, F., Lou, X. Y. & Sun, L. F. An improved chaotic fruit fly optimization based on a mutation strategy for simultaneous feature selection and parameter optimization for SVM and its applications. PLoS ONE 12 , e0173516. https://doi.org/10.1371/journal.pone.0173516 (2017).

Peng, C.-Y.J., Lee, K. L. & Ingersoll, G. M. An introduction to logistic regression analysis and reporting. J. Educ. Res. 96 , 3–14. https://doi.org/10.1080/00220670209598786 (2002).

Park, H.-A. An introduction to logistic regression: From basic concepts to interpretation with particular attention to nursing domain. J. Korean Acad. Nurs. 43 , 154. https://doi.org/10.4040/jkan.2013.43.2.154 (2013).

Article   PubMed   Google Scholar  

Chen, H., Hu, S., Hua, R. & Zhao, X. Improved Naive Bayes classification algorithm for traffic risk management. EURASIP J. Adv. Signal Process. https://doi.org/10.1186/s13634-021-00742-6 (2021).

Shen, J. & Fang, H. Human activity recognition using gaussian Naïve Bayes algorithm in smart home. J. Phys. 1631 , 012059. https://doi.org/10.1088/1742-6596/1631/1/012059 (2020).

Gramegna, A. & Giudici, P. SHAP and LIME: An evaluation of discriminative power in credit risk. Front. Artif. Intell. https://doi.org/10.3389/frai.2021.752558 (2021).

Zaremba, L., Zaremba, C. S. & Suchenek, M. Modification of shapley value and its implementation in decision making. Found. Manag. 9 , 257–272. https://doi.org/10.1515/fman-2017-0020 (2017).

Krishnan, S. R. et al. Smart water resource management using artificial intelligence;a review. Sustainability https://doi.org/10.3390/su142013384 (2022).

Download references

Acknowledgements

Acknowledgements should be brief, and should not include thanks to anonymous referees and editors, or effusive comments. Grant or contribution numbers may be acknowledged.

Author information

Authors and affiliations.

School of Computer Science Engineering and Information Systems, Vellore Institute of Technology, Vellore, 632014, India

M. K. Nallakaruppan, M. Lawanya Shri & Sweta Bhattacharya

Department of Computer Science, Loyola College, Chennai, Tamil Nadu, 600034, India

E. Gangadevi

Shiv Nadar University, Delhi-NCR, 201314, India

Balamurugan Balusamy

School of Built Environment, Engineering and Computing, Leeds Beckett University, Leeds, LS13HE, UK

• Shitharth Selvarajan

Department of Computer Science, Kebri Dehar University, Kebri Dehar, Ethiopia

You can also search for this author in PubMed   Google Scholar

Contributions

All authors contributed equally in this research work.

Corresponding author

Correspondence to Shitharth Selvarajan .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Nallakaruppan, M.K., Gangadevi, E., Shri, M.L. et al. Reliable water quality prediction and parametric analysis using explainable AI models. Sci Rep 14 , 7520 (2024). https://doi.org/10.1038/s41598-024-56775-y

Download citation

Received : 21 October 2023

Accepted : 11 March 2024

Published : 29 March 2024

DOI : https://doi.org/10.1038/s41598-024-56775-y

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Water Quality
• Explainable AI
• Water Quality Prediction

This article is cited by

Erabilnet: enhanced residual attention with bidirectional long short-term memory.

• Koteeswaran Seerangan
• Malarvizhi Nandagopal

Scientific Reports (2024)

Groundwater quality assessment using machine learning models: a comprehensive study on the industrial corridor of a semi-arid region

• Loganathan Krishnamoorthy
• Vignesh Rajkumar Lakshmanan

Environmental Science and Pollution Research (2024)

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Information

• Author Services

Initiatives

You are accessing a machine-readable page. In order to be human-readable, please install an RSS reader.

All articles published by MDPI are made immediately available worldwide under an open access license. No special permission is required to reuse all or part of the article published by MDPI, including figures and tables. For articles published under an open access Creative Common CC BY license, any part of the article may be reused without permission provided that the original article is clearly cited. For more information, please refer to https://www.mdpi.com/openaccess .

Feature papers represent the most advanced research with significant potential for high impact in the field. A Feature Paper should be a substantial original Article that involves several techniques or approaches, provides an outlook for future research directions and describes possible research applications.

Feature papers are submitted upon individual invitation or recommendation by the scientific editors and must receive positive feedback from the reviewers.

Editor’s Choice articles are based on recommendations by the scientific editors of MDPI journals from around the world. Editors select a small number of articles recently published in the journal that they believe will be particularly interesting to readers, or important in the respective research area. The aim is to provide a snapshot of some of the most exciting work published in the various research areas of the journal.

Original Submission Date Received: .

• Active Journals
• Find a Journal
• Proceedings Series
• For Authors
• For Reviewers
• For Editors
• For Librarians
• For Publishers
• For Societies
• For Conference Organizers
• Open Access Policy
• Institutional Open Access Program
• Special Issues Guidelines
• Editorial Process
• Research and Publication Ethics
• Article Processing Charges
• Testimonials
• Preprints.org
• SciProfiles
• Encyclopedia

Article Menu

• Subscribe SciFeed
• Google Scholar
• on Google Scholar
• Table of Contents

Find support for a specific problem in the support section of our website.

Please let us know what you think of our products and services.

Visit our dedicated information section to learn more about MDPI.

JSmol Viewer

A systematic review of water resources assessment at a large river basin scale: case of the major river basins in madagascar.

1. Introduction

2. materials and methods.

• Current approaches for water resources assessment in a large river basin;
• Use of Soil and Water Assessment Tool (SWAT) models and multiple General Circulation Models (GCMs) for water resources assessment;
• Application of inter- and intra-basin approaches to water resources assessment.

3. Results and Discussion

3.1. necessity of water resources assessment at a large river basin scale in madagascar, 3.2. current approaches for water resources assessment at a large river basin, 3.3. use of multiple gcms and swat models for water resources assessment.

• SWt—the final soil water content (mm);
• SWo—the initial soil water content (mm);
• Rday—the amount of precipitation in a day (mm);
• Qsurf—the amount of surface runoff in a day (mm);
• Ea—the amount of evapotranspiration in a day (mm);
• Wseep—the amount of water entering the vadose zone in a day (mm);
• Qgw—the amount of return flow in a day (mm).

3.4. Application of Inter- and Intra-Basin Approaches to Water Resources Assessment

3.5. combination of multiple gcms, hydrological models, and inter- and intra-basin analysis, 4. conclusions and recommendations, author contributions, institutional review board statement, informed consent statement, data availability statement, acknowledgments, conflicts of interest.

• Boretti, A.; Rosa, L. Reassessing the projections of the world water development report. NPJ Clean Water 2019 , 1 , 15. [ Google Scholar ] [ CrossRef ]
• McNally, A.; Verdin, K.; Harrison, L.; Getirana, A.; Jacob, J.; Shukla, S.; Arsenault, K.; Peters-Lidard, C.; Verdin, J.P. Acute water-scarcity monitoring for Africa. Water 2019 , 11 , 1968. [ Google Scholar ] [ CrossRef ]
• Vallee, D.; Margat, J.; Eliasson, A.; Hoogeveen, J. Review of World Water Resources by Country ; Food and Agricultural Organization of the United Nations: Rome, Italy, 2003. [ Google Scholar ]
• Manungufala, T. Water Scarcity: Classification, Measurement and Management. In Clean Water and Sanitation. Encyclopedia of the UN Sustainable Development Goals ; Leal Filho, W., Azul, A.M., Brandli, L., Lange Salvia, A., Wall, T., Eds.; Springer: Cham, Switzerland, 2021. [ Google Scholar ] [ CrossRef ]
• Purvis, L.; Dinar, A. Are intra-and inter-basin water transfers a sustainable policy intervention for addressing water scarcity? Water Secur. 2020 , 9 , 100058. [ Google Scholar ] [ CrossRef ]
• Wada, Y.; Flörke, M.; Hanasaki, N.; Eisner, S.; Fischer, G.; Tramberend, S.; Satoh, Y.; van Vliet, M.T.H.; Yillia, P.; Ringler, C.; et al. Modeling global water use for the 21st century: The Water Futures and Solutions (WFaS) initiative and its approaches. Geosci. Model Dev. 2016 , 9 , 175–222. [ Google Scholar ] [ CrossRef ]
• Schewe, J.; Heinke, J.; Gerten, D.; Haddeland, I.; Arnell, N.W.; Clark, D.B.; Dankers, R.; Eisner, S.; Fekete, B.M.; Colón-González, F.J.; et al. Multimodel assessment of water scarcity under climate change. Proc. Natl. Acad. Sci. USA 2014 , 111 , 3245–3250. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Falkenmark, M.; Anders, B.; Anders, J.; Jan, L.; Manfred, M.; Håkan, T. On the Verge of a New Water Scarcity: A Call for Good Governance and Hyman Ingenuity ; International Water Institute (SIWI): Stockholm, Sweden, 2007. [ Google Scholar ]
• Elimelech, M. The global challenge for adequate and safe water. J. Water Supply: Res. Technol. AQUA 2006 , 55 , 3–10. [ Google Scholar ] [ CrossRef ]
• Munia, H.; Guillaume, J.H.A.; Mirumachi, N.; Porkka, M.; Yoshihide, W.; Kummu, M. Water stress in global transboundary river basins: Significance of upstream water use on downstream stress. Environ. Res. Lett. 2016 , 11 , 014002. [ Google Scholar ] [ CrossRef ]
• UN Water. UN World Water Development Report, Nature-Based Solutions for Water ; UNESCO: Paris, France, 2018. [ Google Scholar ]
• Falkenmark, M. Planning of Africa’s land/water future: Hard or soft landing? Ambio 2022 , 51 , 9–12. [ Google Scholar ] [ CrossRef ]
• Rakotoarimanana, Z.H.; Ishidaira, H. Analysis of River Basin Management in Madagascar and Lessons Learned from Japan. Water 2022 , 14 , 449. [ Google Scholar ]
• Cesar, C.M.; Crestaz, E.; Cimmarrusti, Y.; Farinosi, F.; Biedler, M.; Amani, A.; Mishra, A. Implementing the Water–Energy–Food–Ecosystems Nexus and Achieving the Sustainable Development Goals ; IWA: London, UK, 2021; ISBN 9789231004735. [ Google Scholar ]
• World Meteorological Organization. State of the Climate in Africa 2019 ; WMO-No. 1253; World Meteorological Organization: Geneva, Switzerland, 2020. [ Google Scholar ]
• Guha-Sapir, D.; Hoyois, P.; Below, R. Annual Disaster Statistical Review 2015: The Numbers and Trends ; Centre for Research on the Epidemiology of Disasters (CRED): Brussels, Belgium, 2016. [ Google Scholar ]
• World Meteorological Organization. Technical Material for Water Resources Assessment ; WMO-No.2; World Meteorological Organization: Geneva, Switzerland, 2012. [ Google Scholar ]
• Batchelor, C.; Moriarty, P.; Laban, P. Using water resources assessments within the EMPOWER IWRM planning cycle. EMPOWERS Work. Pap. 2005 , 5 , 30. [ Google Scholar ]
• Martin, E.; Gascoin, S.; Grusson, Y.; Murgue, C.; Bardeau, M.; Anctil, F.; Ferrant, S.; Lardy, R.; Le Moigne, P.; Leenhardt, D.; et al. On the use of hydrological models and satellite data to study the water budget of river basins affected by human activities: Examples from the Garonne basin of France. In Remote Sensing and Water Resources ; Springer: Cham, Switzerland, 2016; pp. 33–57. [ Google Scholar ]
• Cao, X.; Cuyuzuzo, C.M.; Saiken, A.; Song, B. A linear additivity water resources assessment indicator by combining water quantity and water quality. Ecol. Indic. 2021 , 121 , 106990. [ Google Scholar ] [ CrossRef ]
• Luo, P.; Sun, Y.; Wang, S.; Wang, S.; Lyu, J.; Zhou, M.; Nakagami, K.; Takara, K.; Nover, D. Historical assessment and future sustainability challenges of Egyptian water resources management. J. Clean. Prod. 2020 , 263 , 121154. [ Google Scholar ] [ CrossRef ]
• Andrade, C.W.L.; Montenegro, S.M.G.L.; Montenegro, A.A.A.; Lima, J.R.D.S.; Srinivasan, R.; Jones, C.A. Climate change impact assessment on water resources under RCP scenarios: A case study in Mundaú River Basin, Northeastern Brazil. Int. J. Climatol. 2021 , 41 , E1045–E1061. [ Google Scholar ] [ CrossRef ]
• Cuceloglu, G.; Abbaspour, K.C.; Ozturk, I. Assessing the water-resources potential of Istanbul by using a soil and water assessment tool (SWAT) hydrological model. Water 2017 , 9 , 814. [ Google Scholar ] [ CrossRef ]
• Lins, H.F. The imperative of water resources assessment. World Meteorol. Organ. WMO Bull. 2008 , 57 , 159. [ Google Scholar ]
• Biancamaria, S.; Bates, P.D.; Boone, A.; Mognard, N.M. Large-scale coupled hydrologic and hydraulic modelling of the Ob river in Siberia. J. Hydrol. 2009 , 379 , 136–150. [ Google Scholar ] [ CrossRef ]
• Chomba, I.C.; Banda, K.E.; Winsemius, H.C.; Chomba, M.J.; Mataa, M.; Ngwenya, V.; Sichingabula, H.M.; Nyambe, I.A.; Ellender, B. A review of coupled hydrologic-hydraulic models for floodplain assessments in Africa: Opportunities and challenges for floodplain wetland management. Hydrology 2021 , 8 , 44. [ Google Scholar ] [ CrossRef ]
• Nunes Correia, F.; Castro Rego, F.; Da Grača Saraiva, M.; Ramos, I. Coupling GIS with hydrologic and hydraulic flood modelling. Water Resour. Manag. 1998 , 12 , 229–249. [ Google Scholar ] [ CrossRef ]
• Pontes, P.R.M.; Fan, F.M.; Fleischmann, A.S.; de Paiva, R.C.D.; Buarque, D.C.; Siqueira, V.A.; Jardim, P.F.; Sorribas, M.V.; Collischonn, W. MGB-IPH model for hydrological and hydraulic simulation of large floodplain river systems coupled with open source GIS. Environ. Model. Softw. 2017 , 94 , 1–20. [ Google Scholar ] [ CrossRef ]
• McKinney, D.C.; Cai, X.; Rosegrant, M.W.; Ringler, C.; Scott, A.C. Modeling Water Resources Management at the Basin Level: Review and Future Directions ; SWIM Paper 6; International Water Management Institute: Colombo, Sri Lanka, 1999. [ Google Scholar ]
• Xu, C.-Y.; Singh, V.P. Review on regional water resources assessment models under stationary and changing climate. Water Resour. Manag. 2004 , 18 , 591–612. [ Google Scholar ] [ CrossRef ]
• Rayburg, S.; Thoms, M. A coupled hydraulic–hydrologic modelling approach to deriving a water balance model for a complex floodplain wetland system. Hydrol. Res. 2009 , 40 , 364–379. [ Google Scholar ] [ CrossRef ]
• Sui, D.Z.; Maggio, R.C. Integrating GIS with hydrological modeling: Practices, problems, and prospects. Comput. Environ. Urban Syst. 1999 , 23 , 33–51. [ Google Scholar ] [ CrossRef ]
• Brown, C.M.; Lund, J.R.; Cai, X.; Reed, P.M.; Zagona, E.A.; Ostfeld, A.; Hall, J.; Characklis, G.W.; Yu, W.; Brekke, L. The future of water resources systems analysis: Toward a scientific framework for sustainable water management. Water Resour. Res. 2015 , 51 , 6110–6124. [ Google Scholar ] [ CrossRef ]
• Nsubuga, F.N.; Namutebi, E.N.; Nsubuga-Ssenfuma, M. Water resources of Uganda: An assessment and review. J. Water Resour. Prot. 2014 , 6 , 1297. [ Google Scholar ] [ CrossRef ]
• Shahin, M. Review and assessment of water resources in the Arab region. Water Int. 1989 , 14 , 206–219. [ Google Scholar ] [ CrossRef ]
• Digna, R.F.; Mohamed, Y.A.; Van Der Zaag, P.; Uhlenbrook, S.; Corzo, G.A. Nile River Basin modelling for water resources management–a literature review. Int. J. River Basin Manag. 2017 , 15 , 39–52. [ Google Scholar ] [ CrossRef ]
• Khan, K.S.; Kunz, R.; Kleijnen, J.; Antes, G. Five steps to conducting a systematic review. J. R. Soc. Med. 2003 , 96 , 118–121. [ Google Scholar ] [ CrossRef ]
• Xiao, Y.; Watson, M. Guidance on conducting a systematic literature review. J. Plan. Educ. Res. 2019 , 39 , 93–112. [ Google Scholar ] [ CrossRef ]
• Mietton, M.; Gunnell, Y.; Andriamitia, J.; Crouzet, C.; Montade, V.; Jouannic, G.; Nicoud, G.; Razafimahefa, R. Neotectonics and pastoralism: How they impact flood regimes in Madagascar’s highlands. Sci. Total Environ. 2020 , 10 , 140633. [ Google Scholar ] [ CrossRef ]
• Bastaraud, A.; Perthame, E.; Rakotondramanga, J.M.; Mahazosaotra, J.; Ravaonindrina, N.; Jambou, R. The impact of rainfall on drinking water quality in Antananarivo, Madagascar. PLoS ONE 2020 , 15 , e0218698. [ Google Scholar ] [ CrossRef ]
• Brosens, L.; Broothaerts, N.; Campforts, B.; Jacobs, L.; Razanamahandry, V.F.; Van Moerbeke, Q.; Bouillon, S.; Razafimbelo, T.; Rafolisy, T.; Govers, G. Under pressure: Rapid lavaka erosion and floodplain sedimentation in central Madagascar. Sci. Total Environ. 2022 , 806 , 150483. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Rabezanahary Tanteliniaina, M.F.; Rahaman, M.H.; Zhai, J. Assessment of the future impact of climate change on the hydrology of the Mangoky River, Madagascar using ANN and SWAT. Water 2021 , 13 , 1239. [ Google Scholar ] [ CrossRef ]
• Andriambeloson, J.A.; Paris, A.; Calmant, S.; Rakotondraompiana, S. Re-initiating depth-discharge monitoring in small-sized ungauged watersheds by combining remote sensing and hydrological modelling: A case study in Madagascar. Hydrol. Sci. J. 2020 , 65 , 2709–2728. [ Google Scholar ] [ CrossRef ]
• Maina, J.; de Moel, H.; Vermaat, J.E.; Bruggemann, J.H.; Guillaume, M.M.; Grove, C.A.; Madin, J.S.; Mertz-Kraus, R.; Zinke, J. Linking coral river runoff proxies with climate variability, hydrology and land-use in Madagascar catchments. Mar. Pollut. Bull. 2012 , 64 , 2047–2059. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Damkjaer, S.; Taylor, R. The measurement of water scarcity: Defining a meaningful indicator. Ambio 2017 , 46 , 513–531. [ Google Scholar ] [ CrossRef ]
• Singh, V.P. Hydrologic modeling: Progress and future directions. Geosci. Lett. 2018 , 5 , 15. [ Google Scholar ] [ CrossRef ]
• Didovets, I.; Krysanova, V.; Hattermann, F.F.; López, M.D.R.R.; Snizhko, S.; Schmied, H.M. Climate change impact on water availability of main river basins in Ukraine. J. Hydrol. Reg. Stud. 2020 , 32 , 100761. [ Google Scholar ] [ CrossRef ]
• Mallen-Cooper, M.; Brenton, P.Z. History, hydrology and hydraulics: Rethinking the ecological management of large rivers. Ecohydrology 2018 , 11 , e1965. [ Google Scholar ] [ CrossRef ]
• Hamududu, B.H.; Hambulo, N. Impacts of climate change on water resources availability in Zambia: Implications for irrigation development. Environ. Dev. Sustain. 2020 , 22 , 2817–2838. [ Google Scholar ] [ CrossRef ]
• Gelebo, A.H.; Kasiviswanathan, K.S.; Khare, D. Assessment of the spatial–temporal distribution of groundwater recharge in data-scarce large-scale African river basin. Environ. Monit. Assess. 2022 , 194 , 157. [ Google Scholar ] [ CrossRef ]
• Pérez-Martín, M.A.; Estrela, T.; Andreu, J.; Ferrer, J. Modeling water resources and river-aquifer interaction in the Júcar River Basin, Spain. Water Resour. Manag. 2014 , 28 , 4337–4358. [ Google Scholar ] [ CrossRef ]
• Guan, X.; Zhang, J.; Elmahdi, A.; Li, X.; Liu, J.; Liu, Y.; Jin, J.; Liu, Y.; Bao, Z.; Liu, C.; et al. The capacity of the hydrological modeling for water resource assessment under the changing environment in semi-arid river basins in China. Water 2019 , 11 , 1328. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Tshimanga, R.M.; Hughes, D.A. Basin-scale performance of a semidistributed rainfall-runoff model for hydrological predictions and water resources assessment of large rivers: The Congo River. Water Resour. Res. 2014 , 50 , 1174–1188. [ Google Scholar ] [ CrossRef ]
• Sibanda, M.; Mutanga, O.; Chimonyo, V.G.P.; Clulow, A.D.; Shoko, C.; Mazvimavi, D.; Dube, T.; Mabhaudhi, T. Application of drone technologies in surface water resources monitoring and assessment: A systematic review of progress, challenges, and opportunities in the global south. Drones 2021 , 5 , 84. [ Google Scholar ] [ CrossRef ]
• Yan, J.; Jia, S.; Lv, A.; Zhu, W. Water resources assessment of China’s transboundary river basins using a machine learning approach. Water Resour. Res. 2019 , 55 , 632–655. [ Google Scholar ] [ CrossRef ]
• Sánchez-Román, R.M.; Folegatti, M.V.; Orellana, A.M.G. Water resources assessment at Piracicaba, Capivari and Jundiaí River Basins: A dynamic systems approach. Water Resour. Manag. 2010 , 24 , 761–773. [ Google Scholar ] [ CrossRef ]
• Kotir, J.H.; Smith, C.; Brown, G.; Marshall, N.; Johnstone, R. A system dynamics simulation model for sustainable water resources management and agricultural development in the Volta River Basin, Ghana. Sci. Total Environ. 2016 , 573 , 444–457. [ Google Scholar ] [ CrossRef ]
• Du, E.; Tian, Y.; Cai, X.; Zheng, Y.; Li, X.; Zheng, C. Exploring spatial heterogeneity and temporal dynamics of human-hydrological interactions in large river basins with intensive agriculture: A tightly coupled, fully integrated modeling approach. J. Hydrol. 2020 , 591 , 125313. [ Google Scholar ] [ CrossRef ]
• Yang, Q.; Almedinger, J.E.; Zhang, X.; Huang, M.; Chen, X.; Leng, G.; Zhou, Y.; Zhao, K.; Asrar, G.R.; Srinivasan, R.; et al. Enhancing SWAT simulation of forest ecosystems for water resource assessment: A case study in the St. Croix River basin. Ecol. Eng. 2018 , 120 , 422–431. [ Google Scholar ] [ CrossRef ]
• Sholagberu, A.T.; Okikiola, F.O.; Bashir, A.; Adeniyi, A.S.; Juliana, I.O.; Muhammad, M.M.; Abdurrasheed, A.S. Performance Evaluation of SWAT-based Model for the Prediction of Potential and Actual Evapotranspiration. Jordan J. Civ. Eng. 2022 , 16 , 1–11. [ Google Scholar ]
• Gebrechorkos, S.H.; Bernhofer, C.; Hülsmann, S. Climate change impact assessment on the hydrology of a large river basin in Ethiopia using a local-scale climate modelling approach. Sci. Total Environ. 2020 , 742 , 140504. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Nematchoua, M.K.; Ricciardi, P.; Orosa, J.A.; Buratti, C. A detailed study of climate change and some vulnerabilities in Indian Ocean: A case of Madagascar island. Sustain. Cities Soc. 2018 , 41 , 886–898. [ Google Scholar ] [ CrossRef ]
• Didovets, I.; Lobanova, A.; Bronstert, A.; Snizhko, S.; Maule, C.F.; Krysanova, V. Assessment of climate change impacts on water resources in three representative Ukrainian catchments using eco-hydrological modelling. Water 2017 , 9 , 204. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Hooshyar, M.; Singh, A.; Wang, D. Interbasin and intrabasin competitions control drainage network density. Geophys. Res. Lett. 2019 , 46 , 661–669. [ Google Scholar ] [ CrossRef ]
• Jackson, R.B.; Carpenter, S.R.; Dahm, C.N.; McKnight, D.M.; Naiman, R.J.; Postel, S.L.; Running, S.W. Water in a changing world. Ecol. Appl. 2001 , 11 , 1027–1045. [ Google Scholar ] [ CrossRef ]
• Lopes, T.R.; Zolin, C.A.; Mingoti, R.; Vendrusculo, L.G.; de Almeida, F.T.; de Souza, A.P.; de Oliveira, R.F.; Paulino, J.; Uliana, E.M. Hydrological regime, water availability and land use/land cover change impact on the water balance in a large agriculture basin in the Southern Brazilian Amazon. J. S. Am. Earth Sci. 2021 , 108 , 103224. [ Google Scholar ] [ CrossRef ]
• Takele, G.S.; Geremew, S.G.; Azage, G.G.; Agizew, N.E. Hydrological modeling in the Upper Blue Nile basin using soil and water analysis tool (SWAT). Modeling Earth Syst. Environ. 2022 , 8 , 277–292. [ Google Scholar ] [ CrossRef ]
• Swain, S.S.; Ashok, M.; Bhabagrahi, S.; Chandranath, C. Water scarcity-risk assessment in data-scarce river basins under decadal climate change using a hydrological modelling approach. J. Hydrol. 2020 , 590 , 125260. [ Google Scholar ] [ CrossRef ]
• Mengistu, D.; Bewket, W.; Dosio, A.; Panitz, H.J. Climate change impacts on water resources in the upper blue nile (Abay) river basin, ethiopia. J. Hydrol. 2021 , 592 , 125614. [ Google Scholar ] [ CrossRef ]
• Shen, Y.; Taikan, O.; Shinjiro, K.; Naota, H.; Nobuyuki, U.; Masashi, K. Projection of future world water resources under SRES scenarios: An integrated assessment. Hydrol. Sci. J. 2014 , 59 , 1775–1793. [ Google Scholar ] [ CrossRef ]
• Almazroui, M.; Islam, M.N.; Saeed, F.; Saeed, S.; Ismail, M.; Ehsan, M.A.; Diallo, I.; O’Brien, E.; Ashfaq, M.; Martínez-Castro, D.; et al. Projected changes in temperature and precipitation over the United States, Central America, and the Caribbean in CMIP6 GCMs. Earth Syst. Environ. 2021 , 5 , 1–24. [ Google Scholar ] [ CrossRef ]
• Gu, H.; Yu, Z.; Wang, G.; Wang, J.; Ju, Q.; Yang, C.; Fan, C. Impact of climate change on hydrological extremes in the Yangtze River Basin, China. Stoch. Environ. Res. Risk Assess. 2015 , 29 , 693–707. [ Google Scholar ] [ CrossRef ]
• Lee, J.; De Gryze, S.S.J. Effect of climate change on field crop production in the Central Valley of California. Clim. Change 2011 , 109 , 335–353. [ Google Scholar ] [ CrossRef ]
• Arnold, J.G.; Srinivasan, R.; Muttiah, R.S.; Williams, J.R. Large area hydrologic modeling and assessment part I: Model development 1. J. Am. Water Resour. Assoc. 1998 , 34 , 73–89. [ Google Scholar ] [ CrossRef ]
• Akoko, G.; Tu, H.L.; Takashi, G.; Tasuku, K. A review of SWAT model application in Africa. Water 2021 , 13 , 1313. [ Google Scholar ] [ CrossRef ]
• Krysanova, V.; White, M. Advances in water resources assessment with SWAT—An overview. Hydrol. Sci. J. 2015 , 60 , 771–783. [ Google Scholar ] [ CrossRef ]
• Mul, M.; Obuobie, E.; Appoh, R.; Kankam-Yeboah, K.; Bekoe-Obeng, E.; Amisigo, B.; Logah, F.Y.; Gnansah, B.; McCartney, M. Water Resources Assessment of the Volta River Basin ; IWMI Working Paper 166; International Water Management Institute (IWMI): Colombo, Sri Lanka, 2015. [ Google Scholar ]
• Eastham, J.; Mpelasoka, F.; Mainuddin, M.; Ticehurst, C.; Dyce, P.; Hodgson, G.; Ali, R.; Kirby, M. Mekong River Basin Water Resources Assessment: Impacts of Climate Change ; CSIRO: Canberra, Australia, 2008. [ Google Scholar ]
• Shrestha, N.K.; Du, X.; Wang, J. Assessing climate change impacts on fresh water resources of the Athabasca River Basin, Canada. Sci. Total Environ. 2017 , 601 , 425–440. [ Google Scholar ] [ CrossRef ]
• Tan, M.L.; Ibrahim, A.L.; Yusop, Z.; Chua, V.P.; Chan, N.W. Climate change impacts under CMIP5 RCP scenarios on water resources of the Kelantan River Basin, Malaysia. Atmos. Res. 2017 , 189 , 1–10. [ Google Scholar ] [ CrossRef ]
• Chen, J.; Gao, C.; Zeng, X.; Xiong, M.; Wang, Y.; Jing, C.; Krysanova, V.; Huang, J.; Zhao, N.; Su, B. Assessing changes of river discharge under global warming of 1.5 C and 2 C in the upper reaches of the Yangtze River Basin: Approach by using multiple-GCMs and hydrological models. Quat. Int. 2017 , 453 , 63–73. [ Google Scholar ] [ CrossRef ]
• Tegegne, G.; Park, D.K.; Kim, Y.-O. Comparison of hydrological models for the assessment of water resources in a data-scarce region, the Upper Blue Nile River Basin. J. Hydrol. Reg. Stud. 2017 , 14 , 49–66. [ Google Scholar ] [ CrossRef ]
• Tegegne, G.; Park, D.K.; Kim, Y.; Kim, Y.-O. Selecting hydrologic modelling approaches for water resource assessment in the Yongdam watershed. J. Hydrol. 2017 , 56 , 155–164. [ Google Scholar ]
• Guiamel, I.A.; Han, S.L. Watershed modelling of the Mindanao River Basin in the Philippines using the SWAT for water resource management. Civ. Eng. J. 2020 , 6 , 626–648. [ Google Scholar ] [ CrossRef ]
• Xie, P.; Zhuo, L.; Yang, X.; Huang, H.; Gao, X.; Wu, P. Spatial-temporal variations in blue and green water resources, water footprints and water scarcities in a large river basin: A case for the Yellow River basin. J. Hydrol. 2020 , 590 , 125222. [ Google Scholar ] [ CrossRef ]
• Zhang, Y.; Ma, J.; Chang, X.; van Wonderen, J.; Yan, L.; Han, J. Water resources assessment in the Minqin Basin: An arid inland river basin under intensive irrigation in northwest China. Environ. Earth Sci. 2012 , 65 , 1831–1839. [ Google Scholar ] [ CrossRef ]
• Davies, B.R.; Thoms, M.; Meador, M. An assessment of the ecological impacts of inter-basin water transfers, and their threats to river basin integrity and conservation. Aquat. Conserv. Mar. Freshw. Ecosyst. 1992 , 2 , 325–349. [ Google Scholar ] [ CrossRef ]
• Wolf, A.T. Shared waters: Conflict and cooperation. Annu. Rev. Environ. Resour. 2007 , 32 , 241–269. [ Google Scholar ] [ CrossRef ]
• Verdhen, A. Intra and inter basin linking of rivers in water resources management. J. Sci. Ind. Res. 2016 , 75 , 150–155. [ Google Scholar ]
• Snaddon, C.D.; Davies, B.R.; Wishart, M.J.; Meador, M.E.; Thoms, M.C. A Global Overview of Inter-Basin Water Transfer Schemes, with an Appraisal of Their Ecological, Socio-Economic and Socio-Political Implications, and Recommendations for Their Management ; Water Research Commission Report No. TT120/00; Water Research Commission: Pretoria, South Africa, 1999. [ Google Scholar ]
• Nasiri-Gheidari, O.; Safar, M.; Farzaneh, A. A robust multi-objective bargaining methodology for inter-basin water resource allocation: A case study. Environ. Sci. Pollut. Res. 2018 , 25 , 2726–2737. [ Google Scholar ] [ CrossRef ]
• Sun, S.; Zhou, X.; Liu, H.; Jiang, Y.; Zhou, H.; Zhang, C.; Fu, G. Unraveling the effect of inter-basin water transfer on reducing water scarcity and its inequality in China. Water Res. 2021 , 194 , 116931. [ Google Scholar ] [ CrossRef ]
• Shumilova, O.; Tockner, K.; Thieme, M.; Koska, A.; Zarfl, C. Global water transfer megaprojects: A potential solution for the water-food-energy nexus? Front. Environ. Sci. 2018 , 6 , 150. [ Google Scholar ] [ CrossRef ]
• Huang, S.; Shah, H.; Naz, B.S.; Shrestha, N.; Mishra, V.; Daggupati, P.; Ghimire, U.; Vetter, T. Impacts of hydrological model calibration on projected hydrological changes under climate change—A multi-model assessment in three large river basins. Clim. Change 2020 , 163 , 1143–1164. [ Google Scholar ] [ CrossRef ]
• American Institute of Aeronautics and Astronautics. Guide for the Verification and Validation of Computational Fluid Dynamics Simulations ; AIAA-G-077-1998; American Institute of Aeronautics and Astronautics: Reston, VA, USA, 1998. [ Google Scholar ]
• Abbaspour, K.C.; Vejdani, M.; Haghighat, S. SWAT CUP calibration and uncertainty programs for SWAT. In Proceedings of the International Congress on Modelling and Simulation (MODSIM’07), Christchurch, New Zealand, 10–13 December 2007; Oxley, L., Kulasiri, D., Eds.; Modelling and Simulation Society of Australia and New Zealand: Melbourne, Australia, 2007; pp. 1603–1609. [ Google Scholar ]
• Solman, S.A.; Bettolli, M.L.; Doyle, M.E.; Olmo, M.E.; Feijoó, M.; Martinez, D.; Blazquez, J.; Balmaceda Huarte, R. Evaluation of multiple downscaling tools for simulating extreme precipitation events over Southeastern South America: A case study approach. Clim. Dyn. 2021 , 57 , 1241–1264. [ Google Scholar ] [ CrossRef ]
• Thompson, J.R.; Green, A.J.; Kingston, D.G.; Gosling, S.N. Assessment of uncertainty in river flow projections for the Mekong River using multiple GCMs and hydrological models. J. Hydrol. 2013 , 486 , 1–30. [ Google Scholar ] [ CrossRef ]
• Wang, H.M.; Chen, J.; Xu, C.Y.; Zhang, J.; Chen, H. A framework to quantify the uncertainty contribution of GCMs over multiple sources in hydrological impacts of climate change. Earth’s Future 2020 , 8 , e2020EF001602. [ Google Scholar ] [ CrossRef ]
• Song, X.; Zhang, J.; Zhan, C.; Xuan, Y.; Ye, M.; Xu, C. Global sensitivity analysis in hydrological modeling: Review of concepts, methods, theoretical framework, and applications. J. Hydrol. 2015 , 523 , 739–757. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Daggupati, P.; Pai, N.; Ale, S.; Douglas-Mankin, K.R.; Zeckoski, R.W.; Jeong, J.; Parajuli, P.B.; Saraswat, D.; Youssef, M.A. A recommended calibration and validation strategy for hydrologic and water quality models. Trans. ASABE 2015 , 58 , 1705–1719. [ Google Scholar ]
• Van Griensven, A. Sensitivity, auto-calibration, uncertainty and model evaluation in SWAT 2005. UNESCO-IHE 2005 , 48 , 25. [ Google Scholar ]

Click here to enlarge figure

Share and Cite

Zy Harifidy, R.; Zy Misa Harivelo, R.; Hiroshi, I.; Jun, M.; Kazuyoshi, S. A Systematic Review of Water Resources Assessment at a Large River Basin Scale: Case of the Major River Basins in Madagascar. Sustainability 2022 , 14 , 12237. https://doi.org/10.3390/su141912237

Zy Harifidy R, Zy Misa Harivelo R, Hiroshi I, Jun M, Kazuyoshi S. A Systematic Review of Water Resources Assessment at a Large River Basin Scale: Case of the Major River Basins in Madagascar. Sustainability . 2022; 14(19):12237. https://doi.org/10.3390/su141912237

Zy Harifidy, Rakotoarimanana, Rakotoarimanana Zy Misa Harivelo, Ishidaira Hiroshi, Magome Jun, and Souma Kazuyoshi. 2022. "A Systematic Review of Water Resources Assessment at a Large River Basin Scale: Case of the Major River Basins in Madagascar" Sustainability 14, no. 19: 12237. https://doi.org/10.3390/su141912237

Article Metrics

Article access statistics, further information, mdpi initiatives, follow mdpi.

Subscribe to receive issue release notifications and newsletters from MDPI journals

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

The PMC website is updating on October 15, 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Sensors (Basel)

A Comprehensive Review on Water Quality Parameters Estimation Using Remote Sensing Techniques

Mohammad haji gholizadeh.

1 Department of Civil Engineering, Florida International University, Miami, FL 33174, USA; ude.uif@200ijahm (M.H.G.); ude.uif@idderl (L.R.)

Assefa M. Melesse

2 Department of Earth and Environment, Florida International University, Miami, FL 33199, USA

Lakshmi Reddi

Remotely sensed data can reinforce the abilities of water resources researchers and decision makers to monitor waterbodies more effectively. Remote sensing techniques have been widely used to measure the qualitative parameters of waterbodies (i.e., suspended sediments, colored dissolved organic matter (CDOM), chlorophyll- a , and pollutants). A large number of different sensors on board various satellites and other platforms, such as airplanes, are currently used to measure the amount of radiation at different wavelengths reflected from the water’s surface. In this review paper, various properties (spectral, spatial and temporal, etc.) of the more commonly employed spaceborne and airborne sensors are tabulated to be used as a sensor selection guide. Furthermore, this paper investigates the commonly used approaches and sensors employed in evaluating and quantifying the eleven water quality parameters. The parameters include: chlorophyll- a (chl- a ), colored dissolved organic matters (CDOM), Secchi disk depth (SDD), turbidity, total suspended sediments (TSS), water temperature (WT), total phosphorus (TP), sea surface salinity (SSS), dissolved oxygen (DO), biochemical oxygen demand (BOD) and chemical oxygen demand (COD).

1. Introduction

Over 40% of the world’s population lives in coastal regions and lake or river shores [ 1 ], and this proportion is increasing. The coastal area of rivers and other waterbodies are among the most sensitive environments. Any changes in these fragile ecosystems due to anthropogenic activities can endanger the habitats of fish and other aquatic organisms. Similarly, the need for sustainable urban water supplies requires that the quality of existing available water resources as well as their watersheds need to be under continuous monitoring. Besides, the level of treatment required for human consumption, agriculture, animal husbandry and industry necessitates an understanding of the quality of source waters. In this way, at the beginning of the twentieth century, the importance of water quality has to be considered more than ever, and the concentration of chemicals in sewage and industrial discharges in waterbodies needs to be taken under more precise control [ 2 , 3 ].

Water quality indicators including physical, chemical, and biological properties are traditionally determined by collecting samples from the field and then analysing the samples in the laboratory. Although this in-situ measurement offers high accuracy, it is a labour intensive and time consuming process, and hence it is not feasible to provide a simultaneous water quality database on a regional scale [ 4 , 5 ]. Moreover, conventional point sampling methods are not easily able to identify the spatial or temporal variations in water quality which is vital for comprehensive assessment and management of waterbodies. Therefore, these difficulties of successive and integrated sampling become a significant obstacle to the monitoring and management of water quality.

With advances in space science and the increasing use of computer applications and increased computing powers over recent decades, remote sensing techniques have become useful tools to achieve this goal. Remote sensing techniques make it possible to monitor and identify large scale regions and waterbodies that suffer from qualitative problems in a more effective and efficient manner. The collection of remotely sensed data occurs in digital form and therefore is easily readable in computer processing. Remote sensing techniques have been in use since the 1970’s and continue to be widely used in water quality assessment in the contemporary world [ 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 ].

Different sensors mounted on satellites and other platforms, such as aeroplanes, measure the amount of radiation at various wavelengths reflected from the water’s surface. These reflections can be used directly or indirectly to detect different water quality indicators, such as total suspended solids (TSS), chlorophyll- a concentration, turbidity, salinity, total phosphorus (TP), Secchi disk depth (SDD), Temperature, pH, Dissolved Organic Carbon (DOC),etc. The spectral characteristics of water and pollutants, which are functions of the hydrological, biological and chemical characteristics of water, etc. [ 19 ], are essential factors in the monitoring and assessment of water quality. The study thus introduces the widely employed spaceborne and airborne sensors in water quality investigations and discusses the utility of remotely sensed techniques in the qualitative assessment of waterbodies. Various properties (spectral, spatial and temporal, etc.) of spaceborne and airborne sensors are tabulated to be used as a sensor selection guide. Finally, based on the literature survey, the study presents a compilation of the various sensors useful in the study of some measurable water quality parameters, and investigates in more detail eleven water quality parameters based on the employed approaches to measuring their concentrations.

2. An Overview of Water Quality Assessment and Remote Sensing

In-situ data collections are only able to represent point estimations of the quality of water conditions in time and space, and obtaining spatial and temporal variations of quality indices in large waterbodies is almost impossible [ 18 ]. Briefly listed below are the most important limitations associated with conventional methods:

• In-situ sampling and measurements of water quality parameters are labor intensive, time consuming, and costly.
• Investigation of the spatial and temporal variations and water quality trends in large waterbodies is almost impossible.
• Monitoring, forecasting, and management of entire waterbodies might be inaccessible, for example due to the topographic situation.
• Accuracy and precision of collected in-situ data can be questionable due to both field-sampling error and laboratory error.

To overcome these limitations, the use of remote sensing in water quality assessment can be a useful tool. For more than four decades, remote sensing has illustrated strong capabilities to monitor and evaluate the quality of inland waters. Many researchers frequently use the visible and near infrared bands of the solar spectrum (mostly from blue to near infrared region) in their investigations to obtain robust correlations between water column reflection (in some cases emission) and physical and biogeochemical constituents, such as transparency, chlorophyll concentration (phytoplankton), and the organic matters and mineral suspended sediments in different waterbodies [ 9 , 10 , 18 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 ]. Although the capabilities of remote sensing to assess water quality are undeniable, this technique alone is not sufficiently precise and must be used in conjunction with traditional sampling methods and field surveying. In other words, to obtain a better insight, an integrated use of remote sensing, in-situ measurements and computer water quality modelling may lead to an increased knowledge of the water quality of water systems. Collaboration between different governmental, federal and private agencies and data sharing is also helpful to increase the data required for regional studies. Kallio [ 31 ] has mentioned four advantages of applying remote sensing in compliance with other water quality monitoring programs as below:

• Gives a synoptic view of the entire waterbody for more effective monitoring of the spatial and temporal variation.
• Makes it possible to have a synchronized view of the water quality in a group of lakes over a vast region.
• Provides a comprehensive historical record of water quality in an area and represents trends over time.
• Prioritizes sampling locations and field surveying times.

Optically active constituents of water that interact with light and change the energy spectrum of reflected solar radiation from waterbodies can be measured using remote sensing [ 18 ]. The components, already enumerated in the first section, constitute the majority of important water quality parameters in surface waters. Other parameters include acidity, chemicals, and pathogens, which do not change the spectral properties of reflected light and have no directly-detectable signals, but which may be interpretable and inferable from those detectable water quality parameters with which strong correlations can be found [ 18 , 31 ].

3. Spaceborne and Airborne Sensors for Water Quality Studies

Observing sensors are divided into two main categories based on the platforms on which they are situated. Airborne sensors are those that are mounted on a platform within the Earth’s atmosphere (i.e., a boat, a balloon, a helicopter, or an aircraft), and spaceborne sensors are carried by a spacecraft or satellite to locations outside of the Earth’s atmosphere. Understanding the properties of these sensors is necessary to choose an appropriate sensor for the objectives of the study. Therefore, various remote sensing satellites ( Table 1 ) and airborne systems ( Table 2 ) commonly used in water quality assessments, along with their spectral properties including spatial resolution, spectral bands, and revisit interval are presented. This tabulated information is helpful when designing water quality assessment studies, and can be used for the selection of appropriate sensors among many other available sensors in the market.

List of the more commonly used spaceborne sensors in water quality assessments.

Specification of the more commonly used airborne sensors in water quality assessments.

Other categories of sensors that have broad applications in oceanographic remote-sensing are microwave radiometers (MWR) and synthetic aperture radar (SAR). Passive microwave radiometers measure energy emitted at sub-millimeter-to-centimeter wavelengths (at frequencies of 1–1000 GHz) known as microwaves. By understanding the physical processes associated with energy emission at these wavelengths, oceanographers can calculate two important water quality parameters, sea surface temperature (SST) and sea surface salinity (SSS). They have also been used for measuring atmospheric and terrestrial radiation, meteorological studies, such as zenith-pointing surface instruments that view the Earth's atmosphere in a region above the stationary instrument. Table 3 shows the characteristics of the more commonly used microwave radiometers in oceanography and water quality studies.

Characteristics of the more commonly used microwave radiometers in oceanography and water quality studies.

Synthetic aperture radar is a form of radar that used to create two or three-dimensional images of objects [ 32 , 33 , 34 ], and can be mounted on either an aircraft or spacecraft. Although SARs are widely used for water pollution detection like oil pollution, ocean topography, wind speed at the sea surface, and regional ice monitoring, they are not very often applied in water quality studies and measuring water quality parameters. In cases where the data from these sensors are combined with other sensors in water quality studies, the results have demonstrated that SARs are only marginally helpful in improving the estimation of water quality parameters; for example in Zhang [ 35 ] study.

4. Water Quality Investigations through Remote Sensing Techniques

Water quality study is the process of determining the chemical, physical and biological characteristics of waterbodies and identifying the possible contamination sources that degrade the quality of water [ 20 ]. Degradation of the quality of water resources may result from waste discharges, pesticides, heavy metals, nutrients, microorganisms, and sediments. Different water quality standards have been developed to aid in checking the extent of water pollution, and consequently to maintain these quality standards. The most commonly measured qualitative parameters of water are detailed in Table 4 .

The most commonly measured qualitative parameters of water by means of remote sensing.

The terminology and rationale for Case 1 and Case 2 water classifications were established by Morel and Prieur [ 16 ] and Gordon and Morel [ 83 ] in their seminal work on the bio-optical basis for ocean color variations. The definition for Case 1 and Case 2 waters was updated by Mobley et al. [ 84 ] as follows:

• Case 1 waters are those waters whose optical properties are determined primarily by phytoplankton and related colored dissolved organic matter (CDOM) and detritus degradation products.
• Case 2 waters are everything else, namely waters whose optical properties are significantly influenced by other constituents such as mineral particles, CDOM, or microbubbles, whose concentrations do not covary with the phytoplankton concentration.

Remote sensing techniques make it possible to have spatial and temporal view of surface water quality parameters and more effectively and efficiently monitor the waterbodies, and quantify water quality issues. Most of the studies have focused on optically active variables, such as chlorophyll- a (chl- a ), total suspended solids (TSS), and turbidity. There are several other important water quality variables such as pH, total nitrogen (TN), ammonia nitrogen (NH 3 -N), nitrate nitrogen (NO 3 -N), and dissolved phosphorus (DP), which existing literature omit. The main reason is due to their weak optical characteristics and low signal noise ratio. However, these parameters are an important part of water quality indices and are a challenging aspect of research in the field of water quality assessment using remote sensing, which should stimulate and motivate scientists in further efforts. In continuing, the study precisely surveys the more commonly employed approaches in estimating the concentration of the eleven water quality parameters. These water quality indicators include chlorophyll- a (chl- a ), colored dissolved organic matters (CDOM), Secchi disk depth (SDD), turbidity, total suspended sediments (TSS), water temperature (WT), total phosphorus (TP), sea surface salinity (SSS), dissolved oxygen (DO), biochemical oxygen demand (BOD) and chemical oxygen demand (COD).

4.1. Chlorophyll-a

Algal blooms, which are often driven by eutrophication phenomena in freshwater, are directly related to chl- a concentration since it is essential for photosynthesis [ 36 ]. Chl- a is used in oxygenic photosynthesis and is found in plants, algae and cyanobacteria. Chl- a is the major indicator of trophic state because it acts as a link between nutrient concentration, particularly phosphorus, and algal production. Chl- a while mainly reflecting green, absorbs most energy from wavelengths of violet-blue and orange-red light, whose reflectance causes chlorophyll to appear green. Obviously, the addition of chl- b besides chl- a extends the spectrum absorption. Low light conditions tends to favor the production, a rather greater ratio, of chl- b to chl- a molecules, thus increasing photosynthetic yield [ 85 ]. Figure 1 shows the absorption spectrum of both chl- a and chl- b pigments. Many researchers have demonstrated that increasing chl- a concentration causes a decrease in the spectral response at short wavelengths, particularly in the blue band [ 86 , 87 , 88 , 89 , 90 , 91 ]. A large number of studies have focused on chl- a concentration measurement using remote sensing, some of which are cited in this review paper.

The Absorption Spectrum of both the chlorophyll- a and the Chlorophyll- b pigments.

Narrow bands of imagery are required for the measurement of chl- a concentration and its spatial and temporal variations within a waterbody [ 92 ]. A review of the literature on the application of empirical approaches using multispectral sensors shows ambiguous results. While estimating chlorophyll using remote sensing techniques is possible and a lot of authors claim to achieve satisfying results using broadband sensors, several studies exist showed that the broad wavelength spectral data available on current satellites (i.e., Landsat, SPOT) do not permit discrimination of chlorophyll in waters with high suspended sediments [ 86 ]. This is especially due to the dominance of the spectral signal from the suspended sediments, more so in highly turbid and eutrophic waters [ 93 , 94 ].

Since the late 1970’s, many researchers developed bio-optical algorithms that initially were designed for oceans to determine water quality parameters. In these waterbodies, phytoplankton and its breakdown products are the sole determinants of optical properties of the water. In Case 1 waters, by employing an empirical model and interpreting the received radiance at different wavelengths, the concentrations of chl- a can be adequately estimated with satellite images [ 83 ]. In Case 1 waters, spectral bands in the blue to green region are appropriate to identify chl- a concentrations with acceptable precisions. However, in Case 2 waters due to the complexity of the constituents in water, the detection of chl- a is a sophisticated task and requires advanced approaches and techniques. In Case 2 waters (inland and coastal waters), the optical properties are determined, additionally to phytoplankton, by a composite of dissolved organic matter from a terrestrial origin, dead particulate organic matter and inorganic particulate matter. Therefore, determination of chl- a concentration is much more complex and less accurate as these constituents are not statistically correlated. Additionally, the simple fact that Gelbstoff absorption often masks the blue-green region in Case 2 waters implies chl- a algorithms developed for Case 1 waters are not applicable to Case 2 waters [ 95 ].

Various visual spectral bands and their ratios are widely used to quantify chl- a . Spectral band ratios can reduce irradiance, atmospheric and air-water surface influences in the remotely sensed signal [ 86 , 96 ]. The prominent scattering-absorption features of chl- a include strong absorption between 450–475 nm (blue) and at 670 nm (red), and reflectance reaches to peak at 550 nm (green) and near 700 nm (NIR). The reflectance peak near 700 nm and its ratio to the reflectance at 670 nm have been used to develop a variety of algorithms to retrieve chl- a in turbid waters [ 23 ]. Gitelson [ 97 ] studied the behavior of the reflectance peak near 700 nm and concluded that the 700 nm reflectance peak was important for the remote sensing of inland and coastal waters, especially for measuring chlorophyll concentration. Han [ 98 ] pointed out that the spectral regions at 630–645 nm, 660–670 nm, 680–687 nm and 700–735 nm were found to be potential regions where the first derivatives can be used to estimate chlorophyll concentration. Dekker et al. [ 99 ] mentioned that the scattering and absorption characteristics of chl- a can be studied when more than one band is used. Hoogenboom et al. [ 100 ] noted that a ratio using an Advanced Visible–Infrared Imaging Spectrometer (AVIRIS) band located near 713 nm along with the band at 667 nm was the most sensitive for chlorophyll retrieval in inland waters. A similar ratio (R674/R705) has been demonstrated to be optimal for inland lakes and rivers [ 101 ]. Table 5 shows some of the more commonly used techniques for the measurement of chl- a concentration.

Remotely measurements of chl- a using various spectral bands and their ratios.

Other significant literature that applied other approaches to the measurement of chl- a are considered hereafter. Alparslan, Coskun and Alganci [ 61 ] measured the concentration of chl- a using all bands of Landsat-5 TM. Ekercin [ 128 ] used Band 1 (445–530 nm), Band 2 (520–610 nm), Band 3 (640–720 nm), and Band 4 (770–880 nm) of IKONOS data to estimate chl- a concentration in Istanbul, Turkey. Also, Nas, Karabork, Ekercin and Berktay [ 67 ] used the visible near-infrared (VNIR) and the shortwave infrared (SWIR) (first four bands 0.52–1.70 µm) of Terra/ASTER and developed a multiple regression between chl- a concentration and spectral reflectance in the Beysehir Lake, Turkey. Shafique, Fulk, Autrey and Flotemersch [ 29 ], using Compact Airborne Spectrographic Imager (CASI) studied the chl- a concentration in the Great Miami River and 80 miles of the Ohio River. They concluded that linear models using the ratio of wavelengths 705/675 nm could describe chl- a concentration. Bhatti, Rundquist, Schalles and Ramirez [ 40 ], using Airborne Imaging Spectroradiometer for Applications (AISA) sensor in the Apalachicola Bay in Florida, USA, found that two bands reflectance ratio R70 Other significant literature that applied other approaches to the measurement of chl- a are considered in following. Alparslan, Coskun and Alganci [ 61 ] measured the concentration of chl- a using all bands of Landsat-5 TM. Ekercin [ 128 ] used Band 1 (445–530 nm), Band 2 (520–610 nm), Band 3 (640–720 nm), and Band 4 (770–880 nm) of IKONOS data to estimate chl- a concentration in Istanbul, Turkey. Also, Nas, Karabork, Ekercin and Berktay [ 67 ] used the visible near-infrared (VNIR) and the shortwave infrared (SWIR) (first four bands 0.52–1.70 µm) of Terra/ASTER and developed a multiple regression between chl- a concentration and spectral reflectance in the Beysehir Lake, Turkey. Shafique, Fulk, Autrey and Flotemersch [ 29 ], using Compact Airborne Spectrographic Imager (CASI) studied the chl- a concentration in the Great Miami River and 80 miles of the Ohio River. They concluded that linear models using the ratio of wavelengths 705/675 nm can describe chl- a concentration. Bhatti, Rundquist, Schalles and Ramirez [ 40 ], using Airborne Imaging Spectroradiometer for Applications (AISA) sensor in the Apalachicola Bay in Florida, USA, found a significant correlation between the two bands reflectance ratio R700/R670 and chl- a concentration. Also, the three band model R750 * (R670 − 1 − R700 − 1 ) was found to be a predictor of chl- a concentration in Case 2 waters. In addition, the logarithmic ratio of ALOS/AVNIR-2 (Band 3/Band 1) was related with chl- a concentration in his study area. Lim and Choi [ 36 ] using Landsat-8/OLI showed that chl- a presented a good correlation with both OLI bands and band ratio, with calculated R values for Bands 2, 3, 4 and band ratio (Band 5/Band 3) as −0.66, −0.70, −0.64, and −0.64, respectively, at a significance level of p < 0.01. ZHANG and HAN [ 129 ] found that OLI bands 1 to 4 and their combinations had good correlation with chl- a concentration. Kim et al. [ 130 ] using Landsat-8/OLI employed Band 2, Band 5, and a ratio of Band 2/Band 4 to measure chl- a concentration. Mannheim et al. [ 131 ] found that the reflectance curve and the baseline from 672 to 742 nm (CHRIS spectral Bands 8–12) shows the best correlation results and the maximal sensitivity to variations of chl- a concentration. Choe et al. [ 132 ] used Moderate Resolution Imaging Spectroradiometer (MODIS), Sea-viewing Wide Field-of-view Sensor (SeaWiFS), Medium Resolution Imaging Spectrometer (MERIS), and RapidEye for the estimation of chl- a concentration in turbid waters using Two-band and Three-band models based on band ratio such as Red and NIR band.

Furthermore, Qi et al. [ 133 ] developed an approach based on Empirical Orthogonal Function (EOF) analysis to estimate chl- a concentration in surface waters of Taihu Lake, the third largest freshwater lake in China. The EOF approach analyzed the spectral variance of normalized Rayleigh-corrected reflectance (Rrc) data at 469, 555, 645, and 859 nm, and subsequently related that variance to chl- a using 28 concurrent MODIS and field measurements. Feng et al. [ 134 ] developed a new empirical chl- a algorithm for the largest freshwater lake of China (Poyang Lake) using a normalized green-red difference index (NGRDI) and atmospherically-corrected Medium Resolution Imaging Spectrometer (MERIS) data.

Reviewing the literature showed that most algorithms to determine the chl- a concentration need a wavelength near 675 nm and another near 700 nm. For example, the positioning of the spectral bands of Landsat/ETM+ is illustrated in Figure 2 . This figure reveals that the bands are not very broad and also not suitably positioned for the detection of chl- a . Other multispectral sensors such as ASTER, IRS-LISS III and SPOT-HRV have similar spectral positioning in the Red/NIR region and therefore one can conclude that neither of these sensors is very appropriate for the detection of chl- a .

Spectral band positioning of Landsat7/ETM+ on ASD spectroradiometer spectrum [ 95 ].

As mentioned, several satellite and airborne imageries can be used for chl- a estimation. Nonetheless, it revealed that the Landsat TM seems to be more appropriate and widely used for chl- a assessment. Temporal coverage and spatial resolution of TM and its easy accessibility can be the main reasons for the selection of this sensor.

4.2. Colored Dissolved Organic Matters (CDOM)

Colored Dissolved Organic Matters, also called gelbstoff and gilvin, consists of naturally occurring, water-soluble, biogenic, heterogeneous organic substances that are yellow to brown in color [ 135 ], which exist in both fresh and saline waters. These compounds are brown and can color the water yellowish brown in high concentrations. Therefore, they are referred to as yellow matter or colored dissolved organic matter (CDOM), and usually with chl- a and TSS dominate the water color.

CDOM absorbance spectrum can be several times and overlaps the chlorophyll absorption and can account for over 50% of the total absorption at 443 nm, which is the wavelength that chlorophyll concentrations are usually measured [ 136 ]. The increase in the CDOM concentration mainly affects the reflectance values in the blue and green region of the spectrum (especially below ~500 nm) and its absorbance increases exponentially with decreasing wavelength. This effect can complicate the use of chl- a retrieval algorithms and phytoplankton production models that are based on remotely sensed ocean color [ 137 ]. Nonetheless, it is reported by Strömbeck and Pierson [ 138 ] that at high CDOM concentrations, absorbance of red light spectrum can be significant.

Remote sensing of CDOM is important in studying aquatic ecology and carbon dynamics [ 18 , 139 ]. Existence of CDOM in rivers, lakes, and oceans affects the water color as seen by many satellite remote sensing instruments, such as MODIS and SeaWiFS [ 140 ]. CDOM also affects the underwater light field and water’s inherent optical properties (IOP). This characteristic determines the water reflectance received by remote sensors. Therefore, inversion of remote sensing data provides an efficient method to estimate CDOM concentration within a large spatial and temporal scale [ 141 , 142 ]. In ocean color studies, CDOM absorption properties, for example, its absorption coefficients at 440 nm, are usually used as a representative of CDOM concentration [ 143 ]. In the algorithms derived from sensors like CZCS, chl- a concentrations were empirically inverted, and accordingly, CDOM can be measured with the assumption that it co-varies with chlorophyll [ 16 , 83 , 144 ]. Hyperspectral measurements with newly developed remote sensing reflectance models [ 145 ] have also been used to estimate CDOM as one of ocean color components, such as EO-1 Hyperion with MIM (Matrix Inversion Method) [ 8 ]. Kutser et al. [ 146 ] also used band ratio of EO-1/ALI Band 2 and Band 3 to estimate CDOM content in lakes of Southern Finland.

Recent approaches of CDOM estimation combine hyperspectral remote sensing data with semi-analytical models, new factors like the bottom effects, and computational techniques to enhance the accuracy of CDOM inversion [ 49 ]. Traditionally, in most water quality monitoring programs, CDOM absorption is referred to as color and PCU color (Platinum-Cobalt Units) and is used to characterize it. In recent studies, CDOM absorption is directly reported as light absorption coefficients at given wavelengths, and these absorption coefficients and PCU colors are closely correlated [ 136 ]. Semi-analytical models have been developed and applied to SeaWiFS (Sea-viewing Wide Field-of-view Sensor) and MODIS (Moderate Resolution Imaging Spectroradiometer), in which CDOM’s absorption coefficients are directly and independently inverted from remote sensing reflectance (Rrs) [ 49 ]. The semi-analytical models are based on the radiative transfer equations as well as the simplification of radiance and underwater light field [ 142 ]. Remote sensing of CDOM in riverine waters and coastal waters is a challenge and subject to large errors compared to oceanic waters because spectral signals of CDOM usually interfere with chlorophyll and suspended sediments [ 54 , 147 ]. In these complex environments, hyperspectral remote sensing present an advantage due to their spectral responses to water inherent optical properties (IOP) and their broad spectrum of narrow bands.

Several studies have confirmed that high spectral resolution (10 nm or better) can improve the estimation of water inherent optical properties (IOP) in coastal water [ 8 , 148 , 149 ]. However, as mentioned, due to the spectral signal interference from chlorophyll, suspended sediments as well as spatial and temporal heterogeneity of riverine and coastal waters, the applicable bands for CDOM measurement are not always at the same wavelengths. Therefore, identification of significant wavelengths out of hundreds of narrow bands of hyperspectral reflectance is a challenging task [ 55 ]. As a solution, first, the dimensionality of hyperspectral data should be reduced through techniques such as band selection, derivative analysis, spectral indices, or hyperspectral transformation [ 122 , 150 , 151 , 152 ]. It is also necessary to calibrate and validate the remotely sensed CDOM concentrations using a shipboard data acquisition approach concurrently with high spatial resolution underwater CDOM observation. Additionally, CDOM is reported to be commonly used as an important indicator for dissolved organic carbon (DOC) dynamics in freshwater and coastal marine ecosystems [ 153 ] and many observations have provided evidence that CDOM is correlated to DOC [ 54 , 154 , 155 , 156 , 157 , 158 ]. Reviewing the literature revealed that most of the studies are based on four sections: underwater CDOM measurements, in situ hyperspectral measurements, water-surface reflected radiance by means of remote sensor on a satellite or an airborne platform, and functional data analysis [ 49 , 153 ]. The literature showed that CDOM could be quantified using visual spectral bands and their ratios, which is as summarized in Table 6 .

Remotely measurements of CDOM using various spectral bands and their ratios.

Furthermore, Taheri Shahraiyni et al. [ 174 ] by using reflectance values at 490, 510, 560, 620, and 885 nm of MERIS data and applying a fuzzy modeling technique, Active Learning Method (ALM), mapped the spatial distribution of CDOM over the southern parts of the Caspian Sea, Iran. A proxy algorithm was reported for remote sensing of CDOM by an absorption coefficient of ocean water, which is a multi-band quasi-analytical algorithm (QAA) developed by Lee, Carder and Arnone [ 141 ]. Further, alternative algorithms such as computer-based discrete modelling methods are developed for remote sensing of CDOM. However, Kishino et al. [ 175 ] expressed that results can be questionable when a neural network model is implemented to measure the CDOM concentration using ASTER data. Johannessen et al. [ 176 ] using SeaWiFS images found out a relationship between ultraviolet (UV) attenuation coefficient (Kd) at 323 nm, 338 nm, and 380 nm and the Rrs(412)/Rrs(555) band ratio.

Researchers use many sensors to assess CDOM, but SeaWIFS and MODIS, because of their coarse spatial resolution, were widely applied in deep waters. Due to the need for high accuracy for large-scale applications, SeaWiFS data are of little use in shallow waters and hyperspectral imagery like EO-1/Hyperion, EO-1/ALI, and ALOS/AVNIR-2 were preferable for these areas. In addition, a majority of researchers have used a high-resolution spectroradiometer in their in situ hyperspectral measurements to validate their quantified results. These in situ measurements include reflectance values collected at a single certain location and often used as an indicative of individual targets. These data are useful in identifying concentrations of components within the water column and can be collected above and below the water surface [ 22 ]. They are also useful for calibration and validation of remotely sensed estimations of water quality parameters.

4.3. Secchi Disk Depth

Secchi depth is an optical property of water strongly related to water constituents present in the waterbodies. The Secchi depth exhibits an inverse correlation with the amount of total suspended solids (TSS) present in the waterbodies. It can be used to study the relative nutrient and solids loading situations [ 177 ]. The most commonly attempted method for the measurement of water transparency is based on light attenuation principles [ 142 ]. The best-known operational estimation of water transparency is the Secchi disk, created by Pietro Angelo Secchi SJ in 1865, and is a circular disk used for clarity measurements in oceans and lakes. The disc mounts on a line and lowers slowly down in the water until the pattern on the disk is no longer visible. This measure is known as the Secchi disk depth (SDD) and is also related to water turbidity. Figure 3 shows two different kinds of Secchi disks.

Two different kinds of Secchi disks [ 183 ].

SDD is a reasonable indicator of trophic conditions (algal abundance) except in highly colored lakes with low chl- a and non-algal turbidity (clay, calcium carbonate) [ 42 ]. The Secchi depth is inversely correlated with the amount of TSS present in the waterbodies. Therefore, remote sensing can be an ideal tool for monitoring water transparency and estimating the SDD. Recently, Lee et al. [ 178 ] introduced a model to estimate the SDD, which unlike the classical model that relies strongly on the beam attenuation coefficient, the new model relies only on the diffuse attenuation coefficient at a wavelength corresponding to the maximum transparency for such interpretations. Many researchers have applied remote sensing for this purpose and have shown in their studies that remote sensing data is well correlated with SDD values [ 179 , 180 , 181 , 182 ].

SDD has a significant correlation with atmospherically corrected satellite radiance [ 184 , 185 , 186 ]. In atmospherically corrected MSS Green band, SDD is related to reflectance just below the surface, incorporating the ratio of backscattering to total scattering coefficients for suspended particles [ 187 ]. The relationship is quite accurate for SDD < 16 m [ 188 ]. Significant algorithms have been developed for SDD using various remote sensing data, like TM [ 27 , 39 , 102 , 189 , 190 ], MSS [ 24 , 25 , 191 , 192 , 193 ], IKONOS [ 28 , 60 , 128 ] and even video data [ 194 ]. Landsat-TM is one of the most frequently used sensors to estimate SDD. Braga et al. [ 195 ] found that SDD was closely correlated with TM data, especially during high tide. Furthermore, highly suitable models were developed for SDD that ranged from 4 to 15 m from TM1 and TM3 satellite radiance [ 186 ]. However, there was an exception research conducted by Lopez-Garcia and Caselles [ 124 ]. They used TM data and reported that SDD did not show significant correlation with any TM bands. SDD can also be quantified from reflected radiance received by the IRS satellite [ 184 ].

There are many established relationships in the literature between Secchi depth and total phosphorus, chl- a , TSS, and CDOM. The existing literature showed that SDD can be quantified using visual spectral bands and various band ratios. Bhatti, Rundquist, Schalles and Ramirez [ 40 ] used ALOS-AVNIR-2 data and found that the Secchi depth was well correlated with reflectance ratio of R750/R560 (NIR/Green). Thiemann and Kaufmann [ 101 ] used HyMap and CASI data for Secchi disk transparency and chlorophyll- a determination in the Mecklenburg Lake District, Germany. They used the area between a base line and the spectrum from 400 to 750 nm and found a good correlation with the in situ measured Secchi disk transparency (SDT). Ekercin [ 128 ] using Band 1 (445–530 nm), Band 2 (520–610 nm), and Band 3 (640–720 nm) of IKONOS data and developed an algorithm for SDD measurements. Mancino, Nolè, Urbano, Amato and Ferrara [ 26 ] developed an equation using TM1 and the TM3/TM2, TM1/TM2, TM2/TM1 ratios, and Powell et al. [ 196 ] suggested a regression equation related to in-situ Secchi disk transparency measurements by using the Blue, Green, and Red bands of TM. In addition, based on Kloiber et al. (2002) study and TM and MSS imagery analysis, some recommendations were made for a Landsat-based procedure of water clarity assessment. Literature also showed that Secchi disk depth can be quantified using visual spectral bands and various band ratios, which are summarized in Table 7 .

Remotely measurements of SDD using various spectral bands and their ratios.

Although several satellite remote sensing systems have been used to measure the SDD, the relatively low cost, temporal coverage, spatial resolution, and data availability of the Landsat system make it particularly most useful data for the assessment of this water quality parameter. Several studies have demonstrated a strong relationship between Landsat Multispectral Scanner (MSS) or Thematic Mapper (TM) data and ground observations of Secchi depth, which are cited earlier.

SDD and chl- a concentrations have been successfully predicted from satellite image data by developing the relationship between in-situ measurements of SDD and chl- a , and the spectral response of the blue, green, red, and near-infrared bands. This approach has been successfully implemented in Minnesota [ 202 ], Wisconsin [ 203 ], and Michigan [ 204 ] to estimate water clarity for inland lakes, where in-situ data is limited.

4.4. Turbidity and Total Suspended Sediments

Water turbidity is an optical property of water, which scatters and absorbs the light rather than transmit it in straight lines. Suspended sediments are responsible for most of the scattering, whereas the absorption is controlled by chl- a and colored dissolved or particulate matter [ 205 ]. As water turbidity is mainly the result of the presence of suspended matter, turbidity measurement has often been used to calculate fluvial suspended sediment concentrations [ 206 ] and is commonly regarded as the opposite of clarity. The level of turbidity or murkiness is entirely dependent on the amount of suspended particles in a sample of water. The more suspended particles, the more difficult for light to travel through the water and therefore, the higher the water’s turbidity. The complex nature of suspended substances in water changes the reflectance of the waterbody and therefore causes variation in color. To this end, interpretation of remotely sensed data just based on the color of water is not adequate and accurate. Turbidity and total suspended matters are considered as important variables in many studies due to their linkage with incoming sunlight that in turn affects photosynthesis for growth of algae and plankton. These parameters are also directly associated with Secchi disk depth.

Remote sensing techniques are widely used to estimate and map the turbidity and concentrations of suspended particles, and to provide their spatial and temporal variations. Theory shows that use of a single band provides a robust and TSM-sensitive algorithm provided the band is chosen appropriately [ 207 ]. Curran et al. [ 208 ] and Novo et al. [ 209 ] showed that single band algorithms may be adopted where TSM increases with increasing reflectance. However, the complex substances in water change the reflectance of the water body and therefore cause variation in colors, and thus, different spectral bands can be used for TSS retrievals [ 207 , 210 , 211 ]. For example, high levels of total suspended solids or the presence of dark-colored humus acids from the decay of vegetation, common in the water of peat bogs, would result in high TSS and turbidity readings (Mark and Stapp, 2003). Therefore, the advantage of using signal band or band ratios can be employed to obtain more accurate results in different concentrations in waterbodies. In the Near-IR and Mid-IR regions, water increasingly absorbs the light and makes it look darker, which varies based on water depth and wavelength. An increase of dissolved inorganic materials in waterbodies causes the peak of visible reflectance to shift from the green region (clearer water) toward the red region of the spectrum. Several studies have also found that the first four bands of Landsat are well correlated with total suspended matters [ 42 , 198 , 212 , 213 ]. However, Ritchie et al. [ 214 ] by in situ studies showed that the most useful range of spectrum for the determination of suspended particles in surface waters was between 700 and 800 nm. The literature showed that turbidity and/or Suspended Sediments can be measured using visual spectral bands and various band ratios, which are as summarized in Table 8 .

Remotely measurements of Turbidity and Total Suspended Sediments using various spectral bands and their ratios.

Furthermore, Ekercin [ 128 ] used Band 1 (445–530 nm), Band 2 (520–610 nm), Band 3 (640–720 nm), and Band 4 (770–880 nm) of IKONOS data and estimated the concentration of TSS in Istanbul, Turkey. Alparslan, Coskun and Alganci [ 61 ] obtained the amount of turbidity from Band 1, Band 2, Band 3, Band 4, Band 5 and Band 7 of Landsat-5 TM Satellite Image. He, Chen, Liu and Chen [ 73 ] used a combination of Landsat TM Bands 2, 3, 6 and 7 to correlate with the in situ turbidity measurements. Also, Sudheer, Chaubey and Garg [ 58 ] suggested that a combination of TM1, TM2, TM3 and TM4 was significant to retrieve suspended sediments information from remote sensing data. Bhatti, Rundquist, Schalles and Ramirez [ 40 ] by using NIR/Green band ratio of ALOS-AVNIR-2 developed a relationship to calculate total suspended matters. Lim and Choi [ 36 ] found that suspended solids was correlated with Bands 2–5 of Landsat-8/OLI, and constructed 3 multiple regression models through single bands of OLI.

Reviewing the literature demonstrated that the Landsat/TM was used much more than other sensors. For rivers and other case studies that need more spectral and spatial resolution, ALOS/AVNIR-2, IKONOS on spaceborne sensors, and CASI and AISA hyperspectral imagery on airborne sensors were used to determine turbidity and suspended matters. The methodology to interpret images and to evaluate the turbidity was also improved from simple linear regression to non-linear multiple regression, principle components analysis (PCA) and neural networks.

4.5. Total Phosphorus

Total phosphorus (TP) studies consist of the measurement of all inorganic, organic and dissolved forms of phosphorus. Phosphates are plant nutrients whose increased quantity helps plants and algae to grow quickly. Total phosphorus can be directly related to chl- a concentration and indirectly related to transparency or water clarity, which is estimated by Secchi depth [ 218 ]. Rivers that flow through various land use activities can include different substances and chemicals like total suspended sediments, nutrients, residential fallout, and others. When a river or a creek passes through an agricultural area, for instance, the phosphorus load may show a higher concentration compared to other parameters present in the surface water. Fertilizer-rich agricultural runoffs and effluents from wastewater treatment plants are the main sources of high phosphorus and nitrogen concentrations in surface waters that threaten many worldwide ecosystems [ 219 ]. Total suspended matters usually act as a carrier for TP and also closely related to Secchi disk transparency with an exponential equation [ 220 ].

The measurement of total phosphorus concentrations in waterbodies is challenging due to the spatial heterogeneity and the labor-intensive collection and testing of required field samples. Remote sensing as a robust tool has already been used successfully to monitor water quality parameters in various scales and areas, although it presents a challenge in estimating phosphorus concentration. Remote estimation of total phosphorus (TP) has been investigated based on its high correlation with optically active constituents [ 69 , 220 , 221 , 222 , 223 , 224 ]. Total phosphorus is not directly measurable by optical instruments, but has a general correlation with other water quality parameters. As mentioned above, TP is closely related to some other parameters like phytoplankton [ 220 , 223 ], turbidity and total suspended matters (TSM), and Secchi disk transparency (SDT) [ 224 ], which is the basis for remote monitoring of TP dynamics [ 225 ]. Multispectral Landsat TM data have been widely used to monitor and map the TP spatial and temporal pattern in different regions [ 69 , 221 , 222 ]. Hyperspectral airborne or spaceborne remote sensing due to its finer diagnostic spectral band(s) provides more potential to detect TP in rivers and small lakes.

Many studies have shown that increasing the TP concentration in waterbodies results in a general tendency of increase in chl- a concentration [ 226 , 227 , 228 , 229 ]. Schindler [ 230 ] showed that 74% of the variability in chl- a concentration among lakes has a direct correlation with the variation of phosphorus concentration. His result suggests that chl- a concentration may play a role as a proxy of phosphorus concentration in waterbodies. In another study conducted by Heiskary and Wilson [ 231 ], the Secchi disk depth was decreased with increasing TP concentration that proved that a proportion of phosphorus can be attached to suspended particles resulted from soil erosion and transferred through river’s downslope. These studies suggested that both chl- a concentration and SDD are closely correlated with TP concentration [ 220 ] and therefore can be used as the potential theoretical parameters for the indirect prediction of TP concentration.

Table 9 shows a number of investigations to measure total phosphorus by applying blue band (0.45–0.51 μm) and green band (0.50–0.60 μm), and integration of red (0.60–0.70 μm) and green (0.50–0.60 μm) ratio from different sensors. Empirical estimations and various statistical regression models were used to correlate phosphorus concentration with other water quality indicators, such as Secchi depth (SD) and chl- a concentration. In addition, Bistani [ 232 ] using EO-1/Hyperion obtained a reflectance determination coefficient of 0.49 from the 467 to 529 nm bands ratio values, from which he derived a polynomial algorithm used to produce a total phosphorus distribution map. Song et al. [ 233 ] studied the correlation between TP and TM1, TM2, TM3, and TM4 from the Landsat 5, and found that each band had a correlation with TP of 0.62, 0.59, 0.55, and 0.51, respectively. Later in another study, Song, Li, Li, Tedesco, Hall and Li [ 68 ] by using the airborne imaging data (AISA), and applying red band (around 690 μm) and NIR spectral region (around 710 μm) estimated the total phosphorus (TP) in three central Indiana water supply reservoirs. Wu, Wu, Qi, Zhang, Huang, Lou and Chen [ 69 ] used a combination of TM1, TM3/TM2, and TM1/TM3 data to correlate chl- a concentration and SD measurements with TP concentration. Also, Alparslan, Coskun and Alganci [ 61 ] using Band 1, Band 2, Band 3, Band 4, Band 5 and Band 7 of Landsat-5 TM Satellite Image obtained the amount of total phosphorus concentration. Lim and Choi [ 36 ] used Bands 2, 3, 4, and 5 of Landsat-8/OLI, and constructed 3 multiple regression models by selecting both single bands and band ratios, and obtained significant correlation coefficients.

Remotely measurements of total phosphorus (TP) using various sensors and blue and green bands, and integration of red and green bands ratio.

Results from studied articles indicate that there is a potential to estimate total phosphorus concentration at different scales using airborne and satellite images. The Landsat/TM was used much more than other sensors for TP assessment in the reviewed literature. As phosphorus does not directly present optically diagnostic signals in water leaving radiance for the water quality remote sensing spectral domain (400–900 nm), thus empirical modeling is considered the most applicable approach for the remote estimation of TP in water column [ 68 , 69 , 235 ]. The literature review also showed that TP has a similar spatial pattern to chl- a and SD concentration due to a high correlation of TP with these parameters. Total phosphorus also was highly related to sediment loadings. However, there is a time lag for phytoplankton to consume TP in reservoirs, which make the relationship between TP and chl- a or SD and total suspended sediments more complicated [ 68 ].

Light reflection from the bottom in shallow waters cannot be very reliable, because it may be a result of the above-water remotely sensed reflectance spectra. Therefore, the TP concentration estimated in shallow water may be questionable and needs to be validated using in situ data. Spatial and temporal distribution algorithms for TP concentration produced from satellite-based observations should also be verified by in situ measurements. These empirical methods provide site-specific predictions of total phosphorus with reasonable accuracy [ 236 ].

4.6. Water Temperature

Water temperature is an important parameter for the physical and biochemical processes occurring within water as well as in air-water interactions because temperature regulates physical, chemical, and biological processes in water. Water temperature also influences the solubility, and thus availability of various chemical constituents in water. Most importantly, this parameter affects dissolved oxygen concentrations in water; as oxygen solubility decreases with increasing water temperature. It is also very important to analyze the temporal variations due to seasonal changes. On the other hand, distribution, transportation, and interaction of some contaminants, such as nutrients have a significant relation with water column temperature.

Thermal infrared bands are able to measure the amount of infrared radiant heat emitted from land surfaces and the radiant temperature of waterbodies that have environmental and economic import. As thermal infrared is emitted from the surface, temperature estimations derived by remote sensing must be evaluated with great care when there are reasons to assume that the water is stratified [ 237 ], which occurs shortly after precipitation or solar warming and in areas influenced by freshwater runoff. In such cases, no relation can be expected between sea surface temperatures and the temperatures found in the water under the surface. Thermal stratification in freely flowing rivers is inherently unstable due to variations in channel shape, and in-stream objects, which cause a turbulent flow regime and can usually be detected in the imagery [ 238 ].

Remote sensing of water temperature in rivers is more complex than in other waterbodies because of their much smaller dimensions and difficulties of determination at the resolution (pixel size) of the thermal-infrared (TIR) data [ 45 ]. Stream and river temperature is crucial especially when dealing with endangered fish populations, which are sensitive to increased water temperature. Sparse sampling in both space and time restricts traditional assessment of water temperature, which is typically measured using a network of in-stream gauges, and records the temporal change at given locations. These gages, located in main streams and rivers, are limited in terms of spatial distribution of river temperatures. The application of remote sensing techniques can be an attractive alternative to measuring and monitoring stream temperatures with determined accuracies and uncertainties. Remotely sensed TIR images could provide reliable measurements of the spatial distribution of the stream and river temperature. Remote measurements of water temperature can be obtained with a sensor that detects thermal radiation (3–5 and 8–14 μm wavebands) emitted from the upper 0.1 mm of the water surface [ 239 , 240 , 241 , 242 ].

The emitted TIR radiation (3–14 μm) is a well-established practice, particularly in oceanography where daily observations of regional and global sea-surface temperature (SST) are made from satellites [ 7 , 243 , 244 , 245 ]. In the terrestrial environment, TIR remote sensing of surface water temperature initially focused on lakes [ 246 , 247 ], and coastal applications such as thermal pollution from cooling water discharge from a power plant [ 248 ]. However, starting in the 1990s, airborne TIR remote sensing has been conducted by government agencies over thousands of kilometers of rivers to monitor water quality, identify sources of cold-water inputs, and to develop spatially referenced river temperature models [ 249 , 250 , 251 ]. Currently, many TIR imaging sensors are available that have multiple spectral bands located at different wavelengths, which make them suitable for water temperature measurements. For the selection of appropriate band or bands, careful consideration on the least amount of instrument noise and atmospheric effects is necessary for accurate calculation of the water temperature. However, multiple bands can be averaged to reduce noise due to atmospheric or sensors differences and provide a better estimate of the actual temperature [ 45 ].

Spaceborne TIR imaging sensors cover greater aerial extents compared to airborne TIR imaging sensors. However, significant differences in their range of pixel sizes, number of bands, revisit times, and sensor sensitivities exist. TIR satellite images are an attractive source of broad-scale data due to their low cost, capability for regional coverage, and revisit times, if they are available for the study time, and also have a suitable pixel size. Airborne sensors with finer pixel size are necessary for smaller waterbodies like rivers, but these images are limited to use over large areas because of the high expense of calibrating and processing. In riverine environment, airborne TIR imaging sensors are widely used for monitoring water temperature. When using airborne data acquisition, it is imperative to consider that these images do not provide a truly synoptic assessment of water temperature at a particular time, if the images are collected consecutively along the river course. Therefore, diurnal changes in water temperature should be considered in planning airborne data collection [ 252 ]. TIR imaging system must also be able to minimize internal drift such that frame-to-frame measurements are consistent. In addition, in the case of frame based TIR imaging sensors, the TIR accounting for radiometric distortion must be considered due to variability in individual detector response and lens optics. These uniformity corrections can be performed internally or during the post processing [ 253 ].

Water temperature is a good indicator of the vertical mixing condition and water mass type, and can be used to estimate primary production and phytoplankton growth rates. Preliminary studies have shown that the application of remote sensing combined with traditional in situ temperature measurements can provide reliable information on temperature zones at a relatively low cost. Many studies have shown the applicability of remote sensing to temperature estimation for rivers and streams. For example, Torgersen, Faux, McIntosh, Poage and Norton [ 251 ] used fine pixel-size (0.2–0.4 m) airborne TIR images to evaluate the accuracy of radiant temperature measurements, and found that the remotely sensed radiant temperature was within 0.5 °C of in-situ measurements. They identified that reflected TIR radiation, vertical thermal stratification in the stream, and thermal boundary-layer effects at the water surface should receive greater attention in the thermal remote sensing of streams. They also concluded that fine pixel-size measurements of stream temperature are useful for studying fine-scale spatial variation and patterns in stream temperature related to hydrological features such as ground-water inputs.

Accurate remote sensing measurement of sea surface temperature (SST) is also vital for weather and climate operational as well as atmosphere studies. Infrared radiometers yield SST to around 0.5 °C precision, though its use is limited in shady zones due to the presence of clouds or fog. Therefore, standard remote sensing practices should be applied to identify and mask these issues out of the used images before one proceeds with the measurement of the water temperature by TIR radiation. Passive microwave techniques are used in cloudy areas with an accuracy limit of about 1.5–2 °C by the relatively large variation of microwave emissivity with surface conditions, such as wind speed [ 254 ]. Addition of active microwave (radar) observations can enhance the precision of passive microwave estimates of SST. Reviewing the literature indicates the use of infrared thermal band for quantifying water temperature, which is as summarized in Table 10 .

Infrared thermal band applications to quantify the water temperature.

4.7. Sea Surface Salinity (SSS)

Salinity and temperature are important factors to identify the density of seawater, and in turn, density is a critical component driving the currents in the oceans. Therefore, salinity is one of the key variables worth considering when monitoring and modeling the circulation in oceans. The role of ocean circulation in moderating the climate is crucial, and thus, sea surface salinity (SSS) is also critical to determine the global water balance, productivity forecast models, as well as evaporation rates. For example, when the salinity is relatively low, the mixed layer will be more stable, and the nutrient pump may be partially inhibited, possibly leading to reduced productivity or a delay in the onset of spring and autumn phytoplankton blooms [ 288 ]. Seasonal and inter-annual variability of sea surface salinity represent limitations on the hydrologic balance and coupled ocean-atmosphere climate models [ 288 , 289 ]. Salinity plays a crucial role in the air-sea exchange of gases.

Precipitation makes the ocean water fresher and less dense, which overlays the salty water below, and this thin layer of sea surface fresh water can spoof the shallow satellite readings. The effect of this phenomenon in the tropical ocean, where heavy rainfalls can create pools of local fresh water, is more sensible. It can increase the stability of the upper layer of the water column and significantly reduce the rates of gas transfer across the pycnocline. Spatial and temporal variations in salinity are greater in inland waters and gulfs because they are strongly influenced by climatic events like precipitation vs. evaporation, seasonal river runoff variations, and exchanges with oceans due to tides and flushing times. Spaceborne and airborne experiments in the 1970s and 80s proved the potential of passive L-band microwave radiometers for the measurement of SSS.

Satellite remote sensors provide more frequent and higher spatial resolution data and also make observations at high latitudes. As the measurement of surface salinity by passive microwave radiometers requires long wavelengths (20–30 cm), accurate estimation of SSS from satellite altitudes would require an enormous antenna, which most satellites could not accommodate [ 290 ]. New interferometric technology has made it possible to overcome such problems with antenna size [ 291 , 292 ]. For instance, the Moisture and Ocean Salinity satellite (SMOS) has been in use to measure SSS and provide synthesized SSS maps with a high accuracy. It employs the Microwave Imaging Radiometer using Aperture Synthesis (MIRAS), as the primary instrument on the SMOS, with a fixed two-dimensional interferometric antenna, which operates over a range of incidence angles and makes it different from the old passive microwave imagers. The SMOS operates in a sun-synchronous orbit at an altitude of 760 km with a three-day repeat cycle [ 291 , 292 , 293 ].

Aquarius is another salinity-related sensor that provides the global view of salinity variability required for climate studies. Aquarius was launched on June 10, 2011 and was a NASA microwave radiometer aboard the Argentine. Aquarius L-band radiometer and a scatterometer instrument combination was designed to provide global salinity maps on a monthly basis with a spatial resolution between 76 and 156 km, a swath width of 390 km, and an accuracy of 0.2 psu. The Aquarius was mounted on SAC-D (Scientific Application Satellite-D) and was designed to provide high precision SSS data and monitor the annual and seasonal variation of the large-scale features of the surface salinity field in the open ocean. Additionally, it had the capabilities of conducting long-standing studies regarding how the oceans respond to climate change and the water cycle. For instance, changes in freshwater input and output to the ocean, associated with precipitation, evaporation, ice melting, and river runoff, could be obtained from monthly SSS maps. In addition, Aquarius data are useful for tracking the formation and movement of huge water masses that regulate ocean circulation and the Earth’s climate (NASA, 2011). The Aquarius instrument successfully achieved its science objectives and completed its primary three-year mission in November 2014. Airborne microwave radiometers, such as the Scanning Low-Frequency Microwave Radiometer (SLFMR) and the Salinity, Temperature, and Roughness Remote Scanner (STARRS), have also been used successfully to map SSS and its variability in estuaries and coastal waters.

Many researchers have also used indirect methods based on, for example, satellite-derived temperature profiles, brightness temperature, and CDOM to determine SSS variability. As the salinity has no direct color signal, it could be rather estimated the color signal dominated by major water constituents and developed relationships, such as: (I) Relationship between salinity, temperature, and brightness temperature [ 64 , 288 , 294 , 295 , 296 , 297 , 298 , 299 , 300 , 301 ]; and (II) Relationship between salinity and CDOM [ 54 , 166 , 302 , 303 ]. Remote sensors on aircraft and satellites offer a means for making detailed SSS measurements over large coastal and ocean areas. Some of these experiments are as listed in Table 11 , based on the used sensor.

Remote sensing of Sea Surface Salinity (SSS) based on the used sensor.

Other notable experiences are performed using European Remote Sensing satellite (ERS) C-band scatterometer [ 304 ]; the first seven bands of MODIS [ 305 ]; TOPEX/Poseidon Microwave Radiometer [ 298 , 306 ], and Cooperative Airborne Radiometer for Ocean and Land Studies (CAROLS) L-Band Radiometer [ 64 ]. Nonetheless, a comparison of the various sensors’ characteristics shows that the airborne ESTAR and SLFMR are more appropriate than other instruments to sea surface salinity measurements. That notwithstanding, SMOS and Aquarius are the most widely used sensors for the remote sensing of salinity.

Salinity-measuring satellites require extensive procedures for internal and external calibration and validation to ensure the quality of the geophysical data. These processes are based on characterization measurements, which are performed initially on the ground before launch and, subsequently, in-flight. The calibration scheme encompasses both the spaceborne instrument and the ground data processing. Once the system has been calibrated, the instrument performance can be verified, and the ocean salinity can be measured with a higher level of confidence [ 307 , 308 ].

4.8. Dissolved Oxygen (DO), Biochemical Oxygen Demand (BOD) and Chemical Oxygen Demand (COD)

Dissolved oxygen (DO) is a crucial water quality parameter that influences the living conditions of all aquatic organisms that require oxygen. The level of DO in waterbodies can be affected by anthropogenic activities and natural occurrences in catchments. Water temperature, the amount of oxygen taken out of the system by respiring and decaying organisms, and the amount of oxygen put back into the system by photosynthesizing plants, stream flow, and aeration are the factors that control the amount of dissolved oxygen in waterbodies. The water temperature highly influences the amount of DO; in other words, less oxygen dissolves in warm water than cold water.

Biochemical Oxygen Demand (BOD) is a measure of the amount of oxygen that bacteria will consume under aerobic conditions while decomposing organic matters. Among the sources of food for water-borne bacteria include among others natural organic detritus and organic waste from waste water treatment plants, failing septic systems, as well as agricultural and urban runoff. By exploiting dissolved oxygen, the bacteria decompose these organic materials resulting in a reduction in the level of DO necessary for supporting aquatic life. A simple means of determining the biochemical oxygen demand is by incubating a sealed sample of water for five days and measuring the loss of oxygen from the beginning to the end of the test. Noteworthy is that the need to dilute the samples prior to incubation stem from the likelihood of the bacteria depleting all the oxygen available in the bottle before the test is complete.

Chemical oxygen demand (COD) is the quantity of matter measured with chemical method that needs to be oxidized in water, especially organic contamination. In the context of chemical oxygen demand, no differentiation exists between biologically available and inert organic matter, and COD is largely a measure of the total amount of oxygen required to oxidize all organic material into carbon dioxide and water. BOD values are always less than COD values, yet measuring the latter take only a few hours while measuring BOD takes five days.

Any discharge of effluent with high BOD levels into a stream or river spontaneously accelerates bacterial growth in the river, which in turn consumes and thus reduces the oxygen levels of the water. One pertinent catastrophe is that the oxygen may diminish to levels that are lethal for most fish and many aquatic insects. However, as the river re-aerates due to atmospheric mixing coupled with algal photosynthesis that adds oxygen to the water, the oxygen levels will slowly increase downstream. Routine methods to measure COD are based on points, and have the time-consuming and laborious disadvantages in obtaining the distribution patterns so that it is difficult to reflect the status of whole region synchronously. This kind of point sampling methods may give accurate measurements, but they are time and money consuming. Further and most importantly, they cannot provide the real-time spatial overview that is necessary for the global assessment and monitoring of water quality [ 90 ]. Satellite remote sensing may provide suitable ways to integrate aquatic data collected from traditional in situ measurements.

A review of the available literature confirmed that no single identified and/or recommended sensors can be used with high confidence to perform an appropriate model to measure the reflectance of water resulting from DO, COD, and BOD. There are some examined statistical techniques to determine the relationship between the satellites estimated reflectance and physicochemical parameters of water. Several water quality models were developed to investigate the relationship between the measured values of DO, BOD, and COD in laboratory and remote sensing reflectance, by establishing linear, exponential, and logarithmic regressions. Also, various bands ratios have been studied to obtain the DO, BOD, and COD distribution maps in order to analyze the spatial and temporal changes of these water quality parameters. However, interpretation of the satellite or airborne images and making authentic relationships between spectral characteristics of images and in situ measurements of DO, BOD, and COD in the aquatic ecosystems are still poorly understood. The most notable studies to estimate the amounts of DO, BOD, and COD are as cited in Table 12 .

Remote sensing of dissolved oxygen (DO), biochemical oxygen demand (BOD), and chemical oxygen demand (COD) based on the used sensor.

Although the results of studied articles indicated that the Landsat/TM was used much more than other sensors to estimate the amount of DO, BOD and COD, this research found relatively low potential and accuracy of current remote sensing techniques for the measurements of DO, BOD and COD values in waterbodies, unless there are enough and adequate ground proofs. In situ measurements of surface water radiation and atmospheric corrections of images are vitally important for both the calibration and validation of remotely sensed data.

Despite the fact that remote sensing can be used to reflect many of water quality parameters, such as Secchi disk depth, chlorophyll concentrations, CDOM, total suspended sediments, and temperature, emphasis should be placed on the fact that this technique cannot substitute the traditional methods. The reason behind this is that some parameters of water quality, like DO, BOD and COD cannot be determined with a high level of confidence by these techniques. However, remote sensing techniques are valuable and important for remote areas where direct access is not easy and where the sum of sampling and chemistry analysis costs is high.

5. Limitations of Remote Sensing for the Assessment of Water Quality Parameters

Remote sensing is a suitable technique to study the spatial and temporal variations of water quality variables. However, a number of important constraints that require precise considerations prior to conducting this technique. Developed models from remote sensing data require adequate calibration, and validation using in situ measurements, and can be used only in the absence of clouds. Moreover, the accuracy of extracted water quality parameters might be debatable for some situations; for instance, Kutser [ 170 ] contends that the densest areas of cyanobacteria blooms in the Baltic Sea are usually undetected by standard satellite products due to atmospheric correction or processing errors.

The spatial, temporal, and spectral resolution limitations of many current optical sensors can confine the application of remotely sensed data to assess water quality. Furthermore, certain key parameters that are not easy to measure directly by optical sensors exist, examples of which include water discharge and vertical distribution of water quality parameters in waterbodies. The cost of hyperspectral or airborne data, as well as the required equipment for in situ hyperspectral measurement, is among the main restrictions of using remote sensing methods for water quality assessment. The optical complexities of inland and coastal waters also narrow the scope of remote sensing application [ 40 ].

The segregation of spectral signatures for chl- a , CDOM, and inorganic suspended matter is not well documented in the literature, which is challenging because of the influence of these parameter on each other. The atmospheric interference also restricts the optical signals coming from waterbodies. In the clear sea waters, the maximum light penetration depth expected is about 55 m near 475 nm [ 330 ], and the majority of the incident energy on the water surface is absorbed, and/or transmitted. On the other hand, when the concentrations of suspended sediment extend to say 400 mg/liter, the penetration depth reduces to only 60 cm. Therefore, a progressively thinner layer of surface water is detectable [ 40 ].

Most of the studies have focused on optically active variables, such as chl- a , CDOM, TSS, and turbidity. However, a number of important water quality variables such as PH, total nitrogen (TN), ammonia nitrogen (NH 3 -N), nitrate nitrogen (NO 3 -N), and dissolved phosphorus (DP), etc. are not well investigated due to their weak optical characteristics and low signal noise ratio. Despite the mentioned limitations, remote sensing is still a useful tool for water quality monitoring.

6. Discussion

Several satellite images could be used for water quality assessments. Nonetheless, Landsat TM (Thematic Mapper) images have been used extensively due to their relatively low cost, temporal coverage and spatial resolution [ 3 , 11 , 12 , 26 , 42 , 58 , 61 , 73 , 86 , 88 , 102 , 103 , 105 , 106 , 109 ]. TM data resided on Landsat-5, a sensor that was operational from 1984 until November 2011, and is considered one of the oldest sensors still used for water quality assessment today [ 39 , 61 , 69 , 213 , 216 , 331 ]. Results from earlier studies referenced in tables indicate that the resolution of Landsat TM is suitable for water quality studies.

Information from the available literature revealed that the Landsat sensors, TM (Thematic Mapper), MSS (Multi-Spectral Scanner), ETM (Enhanced Thematic Mapper), and OLI (Operational Land Imager) have been used fairly successfully to measure most of the important water quality parameters, such as chlorophyll- a , Secchi disk depth, Total phosphorus, Total Suspended matters, Turbidity, Dissolved Oxygen, Biochemical Oxygen Demand, and Chemical Oxygen Demand [ 10 , 24 , 39 , 42 , 61 , 75 , 130 , 196 ]. Nonetheless, the use of Landsat data for measuring water quality characteristics has important limitations. The repeat cycle of 16 days imposes major limitations on intra-seasonal monitoring, especially in areas characterized by frequent cloud cover. The water quality parameter characteristics must be related to an “inherent optical property” (IOP) that can be measured by the satellite sensor [ 42 ]. For instance, Kloiber, Brezonik and Bauer [ 24 ] related Secchi disk transparency (SDT) to the radiance measured by Landsat TM and MSS in several spectral bands. Some other potential sources of error due to varying atmospheric conditions are mentioned for Landsat. For example, Brezonik, Menken and Bauer [ 42 ] noted that the measurements of the radiance at the TM sensor are not calibrated for the intensity of incoming solar radiation, which varies with latitude, season, and time of day. Atmospheric interference can be significant over waterbodies because atmospheric haze scatters light (especially blue wavelengths), and the potential for such interference increases as reflected radiance from the water decreases; thus, lake waters with high clarity and high algae and CDOM are most affected. These restrictions may apply to other sensors with similar characteristics to Landsat sensors.

Although, spaceborne hyperspectral sensors with low spatial resolutions, such as MERIS and MODIS have been used for many years to monitor coastal and large inland waterbodies, currently, there are not many hyperspectral sensors in space with suitable spatial resolution capable of monitoring small lakes and ponds. Due to its spatial resolution of 30 m, the hyperspectral imaging spectrometer Hyperion on the EO-1 platform, launched in November 2000, gave rise to new possibilities of operational monitoring [ 332 ]. The Compact High Resolution Imaging Spectrometer (CHRIS) on the platform PROBA with a ground resolution of 18 m complies with these requirements as well. However, since it is a technology demonstrator, it is not in operational use.

Spaceborne and airborne remote sensing and their characteristics, advantages, and disadvantages were discussed previously in this paper. Different considerations of a project, such as required spatial and spectral resolution, geographic coverage area, and project budget determine the preference of one sensor or another. Table 13 represents a summarized comparison of the previously discussed issues related to spaceborne and airborne sensors, where various parameters of these sensors are compared.

Comparison between spaceborne and airborne sensors.

Morel & Gordon [ 333 ] distinguished three different approaches for estimating concentration of water quality parameters; empirical, semi-empirical, and physical or analytical approach. Empirical approaches seek statistical relationships between spectral bands or band combinations and the measured water parameters, without including knowledge of spectral characteristics of the constituents or any physical explanation of the relationship. Semi-empirical methods utilize the physical and spectral information (e.g. absorption features) to develop the algorithms, which are then correlated to the measured constituents. The statistical coefficients are typically bound to the specific region and time of calibration. Analytical approaches determine the constituents’ concentration by modeling the reflectance of surface water and utilizing the inherent and apparent optical characteristics. However, the semi-analytical approaches use simplified analytical models.

The empirical approaches are easy to implement and require less mathematical skills and computation time. Nonetheless, these methods are not suitable for parameters which do not represent distinctive absorption features, like CDOM and to a certain extent for suspended matters [ 334 ]. The analytical approaches can simultaneously determine all constituents of water if the inherent properties of the parameters are well known and large amounts of in situ data are accessible.

7. Conclusions and Recommendation

By increasing the anthropogenic activities and industrial development, water quality has dramatically degraded. Remote sensing and GIS techniques in conjunction with traditional in-situ sampling are the most effective, cheaper and more reliable tools for monitoring water quality parameters in various waterbodies (lakes, rivers, ground water, etc.). From the available literature, one pertinent deduction is that various space-borne and airborne sensors can measure water quality parameters with reliable precision. Newly developed hyperspectral satellite imageries, which can simultaneously record up to 200 spectral channels, such as the Hyperspectral Imager for the Coastal Ocean (HICO), are much more powerful systems for detecting water quality parameters. Also, hyperspectral airborne sensors have greater potential because of their simultaneous collection of narrower and contiguous bands that allow various parameters of water quality to be measured and monitored. Therefore, monitoring and assessing water quality issues through remotely sensed data can result in effective management of water resources. However, few managerial decisions rely on remote sensing-derived water quality evaluations. Instead, current methods for measuring water quality focus on periodic (boat-based) or continuous (ship-based or buoy-based) monitoring models. To best realize the full application potential of remote sensing technologies, an open and effective dialogue is needed between scientists, policy makers, environmental managers, and stakeholders at federal, state, and local level. Results from an internal US Environmental Protection Agency qualitative survey performed by Schaeffer et al. [ 335 ] were used to determine perceptions regarding the use of satellite remote sensing for water quality monitoring to begin understanding why management decisions do not typically rely on satellite-derived water quality products. They also pointed out that difficulties in developing solutions to clarify the perceptions of environmental managers, identified 22 years ago by Specter and Gayle [ 336 ], still exist today.

In most cases, managers and policy makers without technical expertise typically lack the knowledge to understand technical descriptions, abilities and limitations of remote sensing techniques. Therefore, it is essential to clarify perceptions of water quality managers, which does not seem to be necessarily simple or readily achievable. It is highly recommended that researchers, who work in the field of optics and remote sensing beyond publishing manuscripts in peer-reviewed journals, continue to communicate more with water resource management agencies on using appropriate available tools to address important monitoring requirements.

As illustrated in this paper, both satellite and airborne remote sensing are useful in assessing the quality of inland waters. Airborne sensors are more flexible tools than spaceborne sensors because of their higher spatial and spectral resolution coupled with their greater number of spectral bands that makes it possible to retrieve the water quality parameters with more accuracy. Airborne sensors are more suitable to monitor smaller waterbodies, such as rivers and their tributaries, ponds, and estuaries, while satellite sensors are more suitable for the observation of larger waterbodies. In this paper, various properties (spectral, spatial and temporal, etc.) of spaceborne and airborne sensors are tabulated to be used as a sensor selection guide in related studies. Furthermore, based on the literature surveyed, this paper compiled a list of sensors that have been used by researchers to measure various water quality parameters, and compares various parameters of spaceborne and airborne sensors.

Due to the need for high accuracy in local-scale and riverine studies, some of the above mentioned sensors, such as SeaWiFS data are of little use. For these cases, the high resolution and/or hyperspectral remote sensing on spaceborne platforms such as EO-1/Hyperion, ALOS AVNOR-2, IKONOS, HICO, and Landsat-8 and airborne platforms CASI, AISA, AVIRIS, HyMap are recommended for use in water quality measurements.

In addition, the recent advances in computer sciences have had a profound influence on the water quality monitoring, resulting in a broader development of the remote sensing technology. Computers can store and analyze the large sets of data generated by most of the Remote Sensing projects. Also, the use of decision support systems (DSS) and Geographical Information Systems (GIS) provide efficient tools for storing, manipulating and analyzing remote sensing data. GIS can enhance the contributions of water quality modelling for practical water quality forecasting, which is essential for sustainable water resources management and development. Therefore, the excellent practicality and interoperability of the RS and GIS techniques will lead the future water quality models towards integration of RS and GIS techniques and the increased use of these technologies in qualitative studies of water resources. Regardless of numerous endeavors reported in the literature, remote sensing techniques utilized to quantify water quality are yet to be adopted on a routine framework. Based on author’s prior knowledge and experience, and the gained information from this literature review, a schematic flowchart of the supposed framework for water quality monitoring and assessment using remote sensing techniques is presented in Figure 4 . Despite the recent development of analytical approaches, empirical and semi-empirical algorithms are still in extensive use due to the complexity of analytical approaches in terms of their theory and calculation difficulties. Improvement of the methodology to interpret images from simple linear regression to multivariate statistical analysis approaches like principle components analysis (PCA) and neural networks will help to make the procedures more accurate and easier to manipulate.

A suggested remote sensing based framework to predict and assessment of water quality variables.

Acknowledgments

The research was funded by Florida International University, Miami, U.S.A. We appreciate the efforts of all researchers who have worked diligently to advance knowledge and improve outcomes of water quality assessment using remote sensing. We also thank the anonymous reviewers and the editor in chief for their constructive comments that helped to improve the article.

Author Contributions

Mohammad Haji Gholizadeh conceptualized the idea and led the writing of this paper. The first draft was written by Mohammad Haji Gholizadeh and then sent to Lakshmi Reddi and Assefa M. Melesse for their comments and edits. Mohammad Haji Gholizadeh and Assefa M. Melesse compiled all the edits and produced the final paper.

Conflicts of Interest

The authors declare no conflict of interest.

Article  

• Volume 16, issue 8
• BG, 16, 1657–1671, 2019
• Peer review
• Related articles

Reviews and syntheses: Dams, water quality and tropical reservoir stratification

Robert scott winton, elisa calamita, bernhard wehrli.

The impact of large dams is a popular topic in environmental science, but the importance of altered water quality as a driver of ecological impacts is often missing from such discussions. This is partly because information on the relationship between dams and water quality is relatively sparse and fragmentary, especially for low-latitude developing countries where dam building is now concentrated. In this paper, we review and synthesize information on the effects of damming on water quality with a special focus on low latitudes. We find that two ultimate physical processes drive most water quality changes: the trapping of sediments and nutrients, and thermal stratification in reservoirs. Since stratification emerges as an important driver and there is ambiguity in the literature regarding the stratification behavior of water bodies in the tropics, we synthesize data and literature on the 54 largest low-latitude reservoirs to assess their mixing behavior using three classification schemes. Direct observations from literature as well as classifications based on climate and/or morphometry suggest that most, if not all, low-latitude reservoirs will stratify on at least a seasonal basis. This finding suggests that low-latitude dams have the potential to discharge cooler, anoxic deep water, which can degrade downstream ecosystems by altering thermal regimes or causing hypoxic stress. Many of these reservoirs are also capable of efficient trapping of sediments and bed load, transforming or destroying downstream ecosystems, such as floodplains and deltas. Water quality impacts imposed by stratification and sediment trapping can be mitigated through a variety of approaches, but implementation often meets physical or financial constraints. The impending construction of thousands of planned low-latitude dams will alter water quality throughout tropical and subtropical rivers. These changes and associated environmental impacts need to be better understood by better baseline data and more sophisticated predictors of reservoir stratification behavior. Improved environmental impact assessments and dam designs have the potential to mitigate both existing and future potential impacts.

• Article (PDF, 1929 KB)
• Supplement (171 KB)
• Article (1929 KB)
• Full-text XML

Winton, R. S., Calamita, E., and Wehrli, B.: Reviews and syntheses: Dams, water quality and tropical reservoir stratification, Biogeosciences, 16, 1657–1671, https://doi.org/10.5194/bg-16-1657-2019, 2019.

As a global dam construction boom transforms the world's low-latitude river systems (Zarfl et al., 2014) there is a serious concern about how competing demands for water, energy and food resources will unfold. The challenge created by dams is not merely that they can limit the availability of water to downstream peoples and ecosystems, but also that the physical and chemical quality of any released water is often altered drastically (Friedl and Wüest, 2002; Kunz et al., 2011). Access to sufficient quality of water is a United Nations Environment Programme sustainable development goal (UNEP, 2016), and yet the potential negative effects of dams on water quality are rarely emphasized in overviews of impacts of dams (Gibson et al., 2017).

Dams are often criticized by ecologists and biogeochemists for fragmenting habitats (Anderson et al., 2018; Winemiller et al., 2016), disrupting floodplain hydrologic cycles (Kingsford and Thomas, 2004; Mumba and Thompson, 2005; Power et al., 1996) and for emitting large amounts of methane (Delsontro et al., 2011). Such impacts act against the promise of carbon-neutral hydropower (Deemer et al., 2016). In contrast, scientists have committed less investigative effort to documenting the potential impacts of dams on water quality. In cases where investigators have synthesized knowledge of water quality impacts (Friedl and Wüest, 2002; Nilsson and Renofalt, 2008; Petts, 1986), the conclusions are inevitably biased towards mid/high latitudes where the bulk of case studies and mitigation efforts have occurred.

While there is much to be learned from more thoroughly studied high-latitude rivers, given the fundamental role played by climate in river and lake functioning, it is important to consider how the low-latitude reservoirs may behave differently. For example, the process of reservoir stratification, which plays a crucial role in driving downstream water quality impacts, is governed to a large degree by local climate. Additionally, tropical aquatic systems are more prone to suffer from oxygen depletion because of lower oxygen solubility and faster organic matter decomposition at higher temperatures (Lewis, 2000). Latitude also plays an important role in considering ecological or physiological responses to altered water quality. Studies focused on cold-water fish species may have little applicability to warmer rivers in the subtropics and tropics.

Low-latitude river systems are experiencing a high rate of new impacts from very large dam projects. A review of the construction history of very large reservoirs of at least 10 km 3 below ± 35 ∘ latitude reveals that few projects were launched between 1987 and 2000, but in the recent decade (2001–2011) low-latitude mega-reservoirs have appeared at a rate of one new project per year (Fig. 1). Given that ongoing and proposed major dam projects are concentrated at low latitudes (Zarfl et al., 2014), a specific review of water quality impacts of dams and the extent to which they are understood and manageable in tropical biomes is needed.

Figure 1 Construction history of the world's 54 largest reservoirs located below ±35 ∘ latitude. Project year of completion data are from the International Commission on Large Dams ( https://www.icold-cigb.net/ , last access: 1 November 2018). Project start data are approximate ( ±1  year) and based on either gray literature source or for some more recent dams, i.e., visual inspection of Google Earth satellite imagery. Grand Ethiopian Renaissance abbreviated as GER. Volume in map legend is in cubic kilometers.

Large dams exert impacts across many dimension; but in this review, we largely ignore the important, but well-covered, impacts of altered hydrologic regimes. Instead, we focus on water quality, while acknowledging that flow and water quality issues are often inextricable. We also disregard the important issue of habitat fragmentation and the many acute impacts on ecosystems and local human populations arising from dam construction activities (i.e., displacement and habitat loss due to inundation). These important topics have been recently reviewed elsewhere (e.g., Winemiller et al., 2016; Anderson et al., 2018).

In order to understand the severity and ubiquity of water quality impacts associated with dams, it is necessary to understand the process of lake stratification, which occurs because density gradients within lake water formed by solar heating of the water surface prevent efficient mixing. By isolating deep reservoir water from surface oxygen, stratification facilitates the development of low-oxygen conditions and a suite of chemical changes that can be passed downstream. To address the outstanding question of whether low-latitude reservoirs are likely to stratify and experience associated chemical water quality changes, we devote a section of this study to predicting reservoir stratification. This analysis includes the largest low-latitude reservoirs and is focused on physical and chemical processes within the reservoirs that may affect downstream water quality.

Finally, we review off-the-shelf efforts to manage or mitigate undesired chemical and ecological effects of dams related to water quality. The management of dam operations to minimize downstream ecological impacts follows the concept of environmental flows (eflows). The primary goal of eflows is to mimic natural hydrologic cycles for downstream ecosystems, which are otherwise impaired by conventional dam-altered hydrology of diminished flood peaks and higher minimum flows. Although restoring hydrology is vitally important to ecological functioning, it does not necessarily solve water quality impacts, which often require different types of management actions. Rather than duplicate the recent eflow reconceptualizations (e.g., Tharme, 2003; Richter, 2009; Olden and Naiman, 2010; O'Keeffe, 2018), we focus our review on dam management efforts that specifically target water quality, which includes both eflow and non-eflow actions.

The act of damming and impounding a river imposes a fundamental physical change upon the river continuum. The river velocity slows as it approaches the dam wall and the created reservoir becomes a lacustrine system. The physical change of damming leads to chemical changes within the reservoir, which alters the physical and chemical water quality, which in turn leads to ecological impacts on downstream rivers and associated wetlands. The best-documented physical, chemical and ecological effects of damming on water quality are summarized in Fig. 2 and described in detail in this section. In each subsection, we begin with a general overview and then specifically consider the available evidence for low-latitude systems.

Figure 2 Conceptual summary of the physical and chemical water quality effects of dams and how they affect aquatic ecology.

2.1  Stratification-related effects

Stratification, i.e., the separation of reservoir waters into stable layers of differing densities, has important consequences for river water downstream of dams. A key to understanding the impacts of dams on river water quality is a precise understanding of the depth of the reservoir thermocline/oxycline relative to spillway or turbine intakes. At many high-head-storage hydropower dams, the turbine intakes are more than 10–20 m deep to preserve generation capacity even under extreme drought conditions. For example, Kariba Dam's intakes pull water from 20 to 25 m below the typical level of the reservoir surface, which roughly coincides with the typical depth of the thermocline. In more turbid tropical reservoirs, thermocline depth can be much shallower, for example at Murum Reservoir in Indonesia where the epilimnion is just 4 to 6 m thick. Unfortunately, the turbine intake depths are not typically reported in dam databases. Furthermore, for many reservoirs, especially those in the tropics, the mixing behavior and therefore the typical depth of the thermocline are not well understood. Collecting the reservoir depth profiles necessary to generate this key information may be more difficult than simply analyzing water chemistry below dams. Dam tail waters with low oxygen or reduced compounds, such as hydrogen sulfide or dissolved methane, are likely to stem from discharged deep water of a stratified reservoir.

2.1.1  Changing thermal regimes

Even at low latitudes where seasonal differences are less than temperate climates, aquatic ecosystems experience water temperatures that fluctuate according to daily and annual thermal regimes (Olden and Naiman, 2010). Hypolimnetic releases of unseasonably cold water represent alterations to a natural regime. Although the difference between surface and deep waters in tropical lakes is typically much less than those at higher latitudes (Lewis, 1996), the differences are often much greater than the relatively subtle temperature shifts of 3–5  ∘ C that have been shown to cause serious impacts (King et al., 1998; Preece and Jones, 2002). For example, at 17 ∘ south of the Equator, Lake Kariba seasonally reaches a 6 to 7  ∘ C difference between surface and deep waters (Magadza, 2010). The ecological impacts of altered thermal regimes have been extensively documented across a range of river systems.

Many aquatic insects are highly sensitive to alterations in thermal regime (Eady et al., 2013; Ward and Stanford, 1982), with specific temperature thresholds required for completion of various life cycle phases (Vannote and Sweeney, 1980). Since macroinvertebrates form an important prey base for fish and other larger organisms, there will be cascading effects when insect life cycles are disrupted. Fish have their own set of thermal requirements, with species often filling specific thermal niches (Coutant, 1987). Altered thermal regimes can shift species distributions and community composition. Development schedules for both fish and insects respond to accumulated daily temperatures above or below a threshold, as well as absolute temperatures (Olden and Naiman, 2010). Fish and insects have both chronic and acute responses to extreme temperatures. A systemic meta-analysis of flow regulation on invertebrates and fish populations by Haxton and Findlay (2008) found that hypolimnetic releases tend to reduce abundance of aquatic species regardless of setting.

There exist several case studies from relatively low latitudes suggesting that tropical and subtropical rivers are susceptible to dam-imposed thermal impacts. The Murray cod has been severely impacted by cold-water pollution from the Dartmouth Dam in Victoria, Australia (Todd et al., 2005) and a variety of native fish species were similarly impacted by the Keepit Dam in New South Wales, Australia (Preece and Jones, 2002). In subtropical China, cold-water dam releases have caused fish spawning to be delayed by several weeks (Zhong and Power, 2015). In tropical Brazil, Sato et al. (2005) tracked disruptions to fish reproductive success 34 km downstream of the Três Marias Dam. In tropical South Africa, researchers monitoring downstream temperature-sensitive fish in regulated and unregulated rivers found that warm water flows promoted fish spawning, whereas flows of 3 to 5  ∘ C cooler hypolimnetic water forced fish to emigrate (King et al., 1998).

2.1.2  Hypoxia

Stratification tends to lead to the deoxygenation of deep reservoir water, because of heterotrophic consumption and a lack of resupply from oxic surface layers. When dam intakes are deeper than the oxycline, hypoxic water can be passed downstream where it is suspected to cause significant ecological harm. Oxygen concentrations below 3.5 to 5 mg L −1 typically trigger escape behavior in higher organisms, whereas only well-adapted organisms survive below 2 mg L −1 (Spoor, 1990). A study of 19 dams in the southeastern United States found that 15 routinely released water with less than 5 mg L −1 of dissolved oxygen and 7 released water with less than 2 mg L −1 of dissolved oxygen (Higgins and Brock, 1999). Hypoxic releases from these dams often lasted for months and the hypoxic water was detectable in some cases for dozens of kilometers downstream. Below the Hume Dam in Australia, researchers found that oxygen concentrations reached an annual minimum of less than 50 % saturation (well under 5 mg L −1 ), while other un-impacted reference streams always had 100 % oxygen saturation (Walker et al., 1978). Researchers recently observed similar hypoxic conditions below the Bakun Dam in Malaysia with less than 5 mg L −1 recorded for more than 150 km downstream (Wera et al., 2019). Although observations and experiments have demonstrated the powerful stress that hypoxia exerts on many fish species (Coble, 1982; Spoor, 1990), there exist few well-documented field studies of dam-induced hypoxia disrupting downstream ecosystems. This is partly because it can be difficult to distinguish the relative importance of dissolved oxygen and other correlated chemical and physical parameters (Hill, 1968). Hypolimnetic dam releases containing low oxygen will necessarily also be colder than surface waters and they may contain toxic levels of ammonia and hydrogen sulfide, so it was not clear which factor was the main driver for the loss of benthic macroinvertebrate diversity documented below a dam of the Guadalupe River in Texas (Young et al., 1976).

Regardless, regulators in the southern US found the threat of hypoxia to be sufficiently serious to mandate that dam tail waters maintain a minimum dissolved oxygen content of 4 to 6 mg L −1 depending on temperature (Higgins and Brock, 1999). These dams in the Tennessee Valley are on the northern fringe of the subtropics ( ∼35 –36 ∘  N), but are relatively warm compared with other reservoirs of the United States. Since oxygen is less soluble in warmer water and gas transfer is driven by the difference between equilibrium and actual concentrations, it follows that low-oxygen stretches downstream of low-latitude dams will suffer from slower oxygen recovery.

In addition to the direct impact imposed by hypoxic reservoir water when it is discharged downstream, anoxic bottom waters will also trigger a suite of anaerobic redox processes within reservoir sediments that exert additional alterations to water quality. Therefore, anoxia can also exert indirect chemical changes and associated ecological impacts. Here we discuss two particularly prevalent processes: phosphorus remobilization and the generation of soluble reduced compounds.

2.1.3  Phosphorus remobilization and eutrophication

Phosphorus (P) is an important macronutrient. Its scarcity or limited bioavailability to primary producers often limits productivity of aquatic systems. Conversely, the addition of dissolved P to aquatic ecosystems often stimulates eutrophication, leading to blooms of algae, phytoplankton or floating macrophytes on water surfaces (Carpenter et al., 1998; Smith, 2003). Typically, eutrophication will occur when P is imported into a system from some external source, but in the case of lakes and reservoirs internal P loading from sediments can also be important. Most P in the aquatic environment is bound to sediment particles where it is relatively unavailable for uptake by biota, but anoxic bottom waters of lakes greatly accelerate internal P loading (Nurnberg, 1984). Iron oxide particles are strong absorbers of dissolved ;, but under anoxic conditions, the iron serves as an electron acceptor and is reduced to a soluble ferrous form. During iron reduction, iron-bound P also becomes soluble and is released into solution where it can build up in hypolimnetic waters. Water rich in P is then either discharged through turbines or mixed with surface waters during periods of destratification. Therefore, sudden increases in bioavailable P can stimulate algal and other aquatic plant growth in the reservoir epilimnion. Theoretically, discharging of P-rich deep water could cause similar blooms in downstream river reaches, but we are not aware of any direct observations of this phenomenon. Typically, nutrient releases in low-latitude contexts are thought of as beneficial to downstream ecosystems because they would counteract the oligotrophication imposed by the dam through sediment trapping (Kunz et al., 2011).

Although dams seem to typically lead to overall reductions in downstream nutrient delivery (see “Oligotrophication”, Sect. 2.2.2), the phenomenon of within-reservoir eutrophication because of internal P loading has been extensively documented in lakes and reservoirs worldwide. In the absence of major anthropogenic nutrient inputs, the eutrophication is typically ephemeral and is abated after several years following reservoir creation. A well-known tropical example is Lake Kariba, the world's largest reservoir by volume. For many years after flooding a 10 % to 15 % of the lake surface was covered by Kariba weed ( Salvinia molesta ), a floating macrophyte. Limnologists attributed these blooms to decomposing organic matter and also gradual P release from inundated soils exposed to an anoxic hypolimnion (Marshall and Junor, 1981).

Indeed, some characteristics of tropical lakes seem to make them especially susceptible to P regeneration from the hypolimnion. The great depth to which mixing occurs (often 50 m or more) during destratification, a product of the mild thermal density gradient between surface and deep water, provides more opportunity to transport deep P back to the surface (Kilham and Kilham, 1990). This has led limnologists to conclude that deep tropical water bodies are more prone to eutrophication compared with their temperate counterparts (Lewis, 2000). There is of course variability within tropical lakes. Those with larger catchment areas tend to receive more sediments and nutrients from their inflowing rivers and are also more prone to eutrophication (Straskraba et al., 1993). These findings together suggest that thermally stratified low-latitude reservoirs run a high risk of experiencing problems of eutrophication because of internal P re-mobilization.

2.1.4  Reduced compounds

Another ecological stressor imposed by hypoxic reservoir water is a high concentration of reduced compounds, such as hydrogen sulfide ( H 2 S ) and reduced iron, which limit the capacity of the downstream river to cope with pollutants. Sufficient dissolved oxygen is not only necessary for the support of most forms of aquatic life, but it is also essential to maintaining oxidative self-purification processes within rivers (Friedl and Wüest, 2002; Petts, 1986). Reduced compounds limit the oxidative capacity of river water by acting as a sink for free dissolved oxygen. The occurrence of H 2 S has been documented in some cases in the tail waters of dams, but the co-occurrence of this stressor with low temperatures and hypoxia make it difficult to attribute the extent to which it causes direct ecological harm (Young et al., 1976). Researchers investigating fish mortality below Greens Ferry Dam in Arkansas, USA, found H 2 S concentrations of 0.1 mg L −1 (Grizzle, 1981), well above the recognized lethal concentrations of 0.013 to 0.045 mg L −1 of H 2 S for fish based on toxicological studies (Smith et al., 1976). Lethal concentrations of ammonia for fish are 0.75 to 3.4 mg L −1 of unionized  NH 3 (Thurston et al., 1983), though we are unaware of specific cases where these thresholds have been surpassed because of dams. At the very least, the presence of reduced compounds at elevated concentrations indicates that an aquatic system is experiencing severe stresses, which, if sustained, will be lethal to most macroscopic biota.

2.2  Sediment trapping

Dams are highly efficient at retaining sediments (Donald et al., 2015; Garnier et al., 2005; Kunz et al., 2011). As rivers approach reservoirs, the flow velocity slows and loses the potential to slide and bounce along sand and gravel, while lost turbulence allows finer sediments to fall out of suspension. Blockage of sediments and coarse material drives two related impact pathways. The first is physical, stemming from the loss of river sediments and bedload that are critical to maintaining the structure of downstream ecosystems (Kondolf, 1997). The second is chemical; the loss of sediment-bound nutrients causes oligotrophication of downstream ecosystems including floodplains and deltas (Van Cappellen and Maavara, 2016).

2.2.1  Altered habitat

The most proximate impact of sediment starvation is the enhancement of erosion downstream of dams from outflows causing channel incision that can degrade within-channel habitats for macroinvertebrates and fish (Kondolf, 1997). Impacts also reach adjacent and distant ecosystems such as floodplains and deltas, which almost universally depend upon rivers to deliver sediments and nutrients to maintain habitat quality and productivity. In addition to sediment/nutrient trapping, dams also dampen seasonal hydrologic peaks, reducing overbank flooding of downstream river reaches. The combination of these two dam effects leads to a major reduction in the delivery of nutrients to floodplains, which represents a fundamental disruption of the flood pulse, affecting the ecological functioning of floodplains (Junk et al., 1989).

River deltas also rely on sediment delivered by floods and damming has led to widespread loss of delta habitats (Giosan et al., 2014). Sediment delivery to the Mekong Delta has already been halved and could drop to 4 % of baseline if all planned dams for the catchment are constructed (Kondolf et al., 2014a). Elsewhere in the tropics, dam construction has been associated with the loss of mangrove habitat, such as at the Volta estuary in Ghana (Rubin et al., 1999). The morphology of the lower Zambezi's floodplains and delta were dramatically transformed by reduced sediment loads associated with the Cahora Bassa megadam in Mozambique (Davies et al., 2001). With diminished sediment delivery and enhanced erosion from rising sea levels, the future of many coastal deltas is precarious, as most of the world's medium and large deltas are not accumulating sediment fast enough to stay above water over the coming century (Giosan et al., 2014).

2.2.2  Oligotrophication

Although the densely populated and industrialized watersheds of the world typically suffer from eutrophication, dam-induced oligotrophication, through sediment and nutrient trapping, can also severely alter the ecological functioning of rivers and their floodplains, deltas and coastal waters. Globally, 12 % to 17 % of global river phosphorus load is trapped behind dams (Maavara et al., 2015); but in specific locations, trapping efficiency can be greater than 90 %, such as at Kariba Dam on the Zambezi River (Kunz et al., 2011) and the Aswan Dam on the Nile (Giosan et al., 2014). Such extreme losses of sediments and nutrients can cause serious acute impacts to downstream ecosystems, though examples are relatively scarce because predam baseline data are not often available.

Most of the best-documented examples of impacts stemming from oligotrophication are from temperate catchments with important and carefully monitored fisheries. For example damming led to the collapse of a valuable salmon fishery in Kootenay Lake, British Columbia, Canada, through oligotrophication (Ashley et al., 1997). The fishery was eventually restored through artificial nutrient additions. Oligotrophication may impose similar ecological impacts in tropical contexts, such as in southern Brazil where an increase in water clarity following the closure of the Eng. Sérgio Motta Dam (Porto Primavera Dam) was associated with a shift in fish communities (Granzotti et al., 2018). Impacts have been perhaps most dramatic in the eastern Mediterranean following the closure of the Aswan Dam. The Lake Nasser reservoir, following its closure in 1969, began capturing the totality of the Nile's famously sediment-rich floodwaters, including some 130 million tons of sediment that had previously reached the Mediterranean Sea. In the subsequent years there was a 95 % drop in phytoplankton biomass and an 80 % drop in fish landings (Halim, 1991). With dams driving rivers toward oligotrophy, and land-use changes such as deforestation and agricultural intensification, causing eutrophication, globally most rivers face some significant change to trophic state.

2.2.3  Elemental ratios

The attention to phosphorus and nitrogen can obscure the importance of other nutrients and their ratios. The element silicon (Si), which is also efficiently sequestered within reservoirs, is an essential nutrient for certain types of phytoplankton. The simultaneous eutrophication and damming of many watersheds has led to decreases in Si to nitrogen ratios, which tends to favor nonsiliceous species over diatoms (Turner et al., 1998). In the river Danube efficient trapping of Si in reservoirs over several decades led to a shift in Black Sea phytoplankton communities (Humborg et al., 1997), coinciding with a crash in an important and productive fishery (Tolmazin, 1985). Turner et al. (1998) document a similar phenomenon at lower latitude in the Mississippi Delta. Decreases in Si loading led to a drop in the abundance of copepods and diatoms relative to the total meso-zooplankton population in the Gulf of Mexico (Turner et al., 1998). These community shifts may have important implications for coastal and estuarine fish communities and the emergence of potential harmful algal blooms.

2.3  Reversibility and propagation of impacts

One way to think of the scope of dam impacts on water quality is in terms of how reversible perturbations to each variable may be. Sediment trapped by a dam may be irreversibly lost from a river and even unregulated downstream tributaries are unlikely to compensate. Temperature and oxygen impacts of dams, in contrast, will happen gradually through exchange with the atmosphere as the river flows. The speed of recovery will depend upon river depth, surface areas, turbulence and other factors that may provide input to predictive models (Langbein and Durum, 1967). Field data from subtropical Australia and tropical Malaysia suggest that, in practice, hypoxia can extend to dozens or hundreds of kilometers downstream of dam walls (Walker et al., 1978; Wera et al., 2019). Where reaeration measures are incorporated into dam operations, hypoxia can be mitigated immediately or within a few kilometers, as was the case in Tennessee, USA (Higgins and Brock, 1999). A study in Colorado, USA, found that thermal effects could be detected for hundreds of kilometers downstream (Holden and Stalnaker, 1975). Regardless of the type of impact, it is clear that downstream tributaries play an important role in returning rivers to conditions that are more “natural” by providing a source of sediment and flow of more appropriate water quality. Water quality impacts of dams are therefore likely to increase and become less reversible when chains of dams are built along the same river channel or on multiple tributaries of a catchment network.

Because the chemical changes in hypoxia and altered thermal regimes both stem from the physical process of reservoir stratification, understanding a reservoir's mixing behavior is an important first step toward predicting the likelihood of water quality impacts. Unfortunately, there exists ambiguity and misinformation in the literature about the mixing behavior of low-latitude reservoirs. To resolve the potential confusion, we review literature on the stratification behavior of tropical water bodies and then conduct an analysis of stratification behavior of the 54 most-voluminous low-latitude reservoirs.

3.1  Stratification in the tropics

For at least one authority on tropical limnology, the fundamental stratification behavior of tropical lakes and reservoirs is clear. Lewis (2000) states that “Tropical lakes are fundamentally warm monomictic …” with only the shallowest failing to stratify at least seasonally, and that periods of destratification are typically predictable events coinciding with cool, rainy and/or windy seasons. Yet, there exists confusion in the literature. For example, the World Commission on Large Dams' technical report states that stratification in low-latitude reservoirs is “uncommon” (McCartney et al., 2000). The authors provide no source supporting this statement, but the conclusion likely stems from the 70-year-old landmark lake classification system (Hutchinson and Loffler, 1956), which, based on very limited field data from equatorial regions, gives the impression that tropical lakes are predominately either oligomictic (mixing irregularly) or polymictic (mixing many times per year). The idea that low-latitude water bodies are fundamentally unpredictable or aseasonal, as well as Hutchinson and Loffler's (1956) approach of classifying lakes without morphometric information critical to understanding lake stability (Boehrer and Schultze, 2008), has been criticized repeatedly over subsequent decades as additional tropical lake studies have been published (Lewis, 1983, 2000, 1973, 1996). And yet, the original misleading classification diagram continues to be faithfully reproduced in contemporary limnology text books (Bengtsson et al., 2012; Wetzel, 2001). Since much of the water quality challenges associated with damming develop from the thermal and/or chemical stratification of reservoirs, we take a critical look at the issue of whether low-latitude reservoirs are likely to stratify predictably for long periods.

3.2  The largest low-latitude reservoirs

To assess the prevalence of prolonged reservoir stratification periods that could affect water quality, we reviewed and synthesized information on the 54 most-voluminous low-latitude reservoirs. Through literature searches, we found descriptions of mixing behavior for 32 of the 54. Authors described nearly all as “monomictic” (having a single well-mixed season, punctuated by a season of stratification). One of these reservoirs was described as meromictic (having a deep layer that does not typically intermix with surface waters) (Zhang et al., 2015). The review indicates that 30 reservoirs stratify regularly for seasons of several months and thus could experience the associated chemical and ecological water quality issues, such as thermal alterations and hypoxia. The two exceptions are Brazilian reservoirs, Três Irmãos and Ilha Solteira, described by Padisak et al. (2000) to be “mostly polymitic” (mixing many times per year). On further investigation, this classification does not appear to be based on direct observations but is a rather general statement of regional reservoir mixing behavior (see Supplement).

We compared our binary stratification classification based on available literature to the results of applying reservoir data to three existing stratification classification schemes. First, we consider the classic Hutchinson and Loffler (1956) classification diagram based on altitude and latitude. Second, we plot the data onto a revised classification diagram for tropical lakes proposed by Lewis (2000) based on reservoir morphometry. Finally we apply the concept of densimetric Froude number which can be used to predict reservoir stratification behavior (Parker et al., 1975) based on morphometry and discharge.

The Hutchinson and Loffler (1956) classification is meant to be applied to “deep” lakes and therefore is not useful for discriminating between stratifying and nonstratifying reservoirs based on depth. It does suggest that all sufficiently deep reservoirs (except those above 3500 m altitude) should stratify. Most of the reservoirs in our data set fall into an “oligomictic” zone, indicating irregular mixing (Fig. 3) when available literature suggests that most would be better described as monomictic, with a predictable season of deeper mixing. This finding reaffirms one of the long-running criticisms of this classification scheme: its overemphasis on oligomixis (Lewis, 1983).

We found that the Lewis (2000) classification system for tropical lakes correctly identified most of the reservoirs in our data set as monomictic; however, six relatively shallow reservoirs known to exhibit seasonal stratification were misclassified into polymictic categories (Fig. 4a). Five of these six lie within a zone labeled “discontinuous polymictic,” which refers to lakes that do not mix on a daily basis, but mix deeply more often than once per year. The literature suggests that these lakes would be better described as “monomictic.” We should note that Lewis's (2000) goals in generating this diagram were to improve upon the Hutchinson and Loffler (1956) diagram for low-latitude regions and to develop a classification system that could be applied to shallow lakes. Lewis (2000) does not mathematically define the boundaries of difference and describes them as “approximate” based on his expert knowledge, so it is not terribly surprising that there appears to be some misclassifications.

Figure 3 The 54 most-voluminous low-latitude reservoirs overlaid onto a lake classification diagram (redrawn from Hutchinson and Loffler, 1956).

Figure 4 Reservoir morphometry and stratification behavior. (a)  Relationships between area and depth for 40 of the 54 world's most-voluminous reservoirs located below ±35 ∘ latitude. Data are from the International Commission on Large Dams ( https://www.icold-cigb.net/ , last access: 1 August 2018) (we excluded 14 reservoirs because of missing surface area data). Stratification behavior classification is synthesized from literature: circle symbols indicate that the reservoir has an extended, predictable season of stratification and/or mixes deeply no more than once per year; triangle symbols refer to two Brazilian reservoirs that authorities suggest are likely to be polymictic, but for which no direct observations exist (see Tundisi, 1990; Padisak, 2000); for reservoirs indicated by cross symbols, no published information on stratification behavior appears in literature searches. Dashed lines and classification labels are approximations proposed by Lewis (2000). (b)  Reservoirs sorted by densimetric Froude number, which is a function of reservoir depth, length, volume and discharge (Parker et al., 1975). The vertical dashed lines at Fr   =1 and Fr   =0.3 indicate the expected boundaries between strongly, weakly and nonstratifying reservoirs (Orlob, 1983). Small dots represent Froude numbers if maximum depth (height of dam wall) is used instead of mean depth as suggested by Ledec and Quintero, 2003. Discharge data is from the Global Runoff Data Centre ( https://www.bafg.de/GRDC , last access: 1 August 2018); five dams were excluded because of missing discharge data.

In a final stratification assessment, we compared the known mixing behavior from literature to calculations of densimetric Froude number ( Fr ) (see Supplement) for the 35 reservoirs for which discharge and surface area data are available (Fig. 4b). We calculated Fr using two different values for depth: first, using mean depth by dividing reservoir volume by area; and second using dam wall height as a proxy for maximum depth. While some authors have suggested that either value for depth can be used (Ledec and Quintero, 2003), our analysis suggests that this choice can have a strong impact on Fr calculations and interpretation. The ratio between mean and maximum depth within our data set ranges from 0.1 to 0.4 with a mean of 0.23. This means that all Fr values could be recalculated to be roughly one-quarter of their value based on mean depth. This is inconsequential for reservoirs with small Fr , but for those with large Fr it can lead to a shift across the classification thresholds of 0.3 and 1. For example, two reservoirs in our data set exceeded the threshold of Fr   =1 to indicate nonstratifying behavior when using mean depth, but they drop down into the weakly stratifying category when maximum depth is used instead. Nine other reservoirs shift from weakly to strongly stratifying. So which value for Fr better reflects reality? It is worth considering that reservoirs are typically quite long and limnologists often break them down into subbasins, separating shallower arms closer to river inflows from deeper zones close to the dam wall. Use of maximum depth for Fr calculations probably better reflects stratification behavior at the dam wall, whereas average depth may better indicate behavior in shallower subbasins that are less likely to stratify strongly. Since the deepest part of a reservoir is at the dam wall and because stratification in this zone is the most relevant to downstream water quality, it is probably most appropriate to use maximum depth in Fr calculations. The two reservoirs described as polymictic by Tundisi (1990) and Padisak et al. (2000) fall into the intermediate category of weakly stratifying (when using mean depth), but three others within this zone are reported to exhibit strong stratification (Deus et al., 2013; Naliato et al., 2009; Selge and Gunkel, 2013). Better candidates for nonstratifying members of this reservoir data set are Yacyretá and Eng. Sérgio Motta; but unfortunately, we could find no description of their mixing behavior in the literature. A field study with depth profiles of these reservoirs could dispel this ambiguity and determine whether all of the largest low-latitude reservoirs stratify on a seasonal basis.

Overall, this exercise of calculating Fr for large low-latitude reservoirs seems to indicate that most, if not all, are likely to stratify. This is an important realization because it points to a significant probability for downstream water quality problems associated with deep-water releases. Furthermore, the bulk of evidence suggests that these reservoirs mix during a predictable season and not irregularly throughout the year or across years, indicating that under normal conditions these reservoirs should be able to stratify continuously for periods of at least a few months.

4.1  Environmental flows

The most developed and implemented approach (or collection of approaches; reviewed by Tharme, 2003) for the mitigation of dam impacts is the environmental flow (eflow), which seeks to adjust dam releases to mimic natural hydrologic patterns. On a seasonal scale, large dams often homogenize the downstream discharge regime. An eflow approach to reservoir management could implement a simulated flood period by releasing some reservoir storage waters during the appropriate season. Although the eflow approach has traditionally focused on the mitigation of ecological problems stemming from disrupted hydrologic regimes, there is a growing realization that water quality (variables such as water temperature, pollutants, nutrients, organic matter, sediments, dissolved oxygen) must be incorporated into the framework (Olden and Naiman, 2010; Rolls et al., 2013). There is already some evidence that eflows successfully improve water quality in practice. In the Tennessee valley, the incorporation of eflows into dam management improved downstream dissolved oxygen (DO) and macroinvertebrate richness (Bednařík et al., 2017). Eflows have also been celebrated for preventing cyanobacteria blooms that had once plagued an estuary in Portugal (Chícharo et al., 2006). These examples illustrate the potential for eflows to solve some water quality impacts created by dams.

Unfortunately, eflows alone will be insufficient in many contexts. For one, flow regulation cannot address the issue of sediment and nutrient trapping without some sort of coupling to a sediment flushing strategy. Second, the issues of oxygen and thermal pollution often persist under eflow scenarios when there is reservoir stratification. Even if eflows effectively simulate the natural hydrologic regime, there is no reason why this should solve water quality problems as long as the water intake position is below the depth of the reservoir thermocline and residence time is not significantly changed. The solution to hypoxic, cold water is not simply more of it; but rather engineers must modify intakes to draw a more desirable water source, or destratification must be achieved. To address the problems of aeration and cold-water pollution, dam managers turn to outflow modification strategies or destratification.

4.2  Aeration

Hypoxia of reservoir tail waters is a common problem imposed by dams. As a result, various management methods for controlling dissolved oxygen content in outflows exist, ranging in cost-effectiveness depending on the characteristics of the dam in question (reviewed by Beutel and Horne, 1999). Options include turbine venting, turbine air injection, surface water pumps, oxygen injection and aerating weirs. As the issue of dam-induced hypoxia has been recognized for many decades, most modern dams incorporate some sort of oxygenation design elements. Where rules strictly regulate dissolved oxygen in the Tennessee Valley, United States, hydropower operators continuously monitor DO in large dam outflows. DO levels are managed by hydropower plant personnel specializing in water quality, aeration and reservoir operations (Higgins and Brock, 1999). However, dams do not always function as designed, especially older constructions in regions with less regulation and oversight; and in such cases, retrofits or adjusted management strategies may be effective. For example, Kunz et al. (2013) suggest that hypoxic releases from Zambia's Itezhi-Tezhi Dam (built in the 1970s) could be mitigated by releasing a mixture of hypolimnetic and epilimnetic waters. This proposed action could help protect the valuable fisheries of the downstream Kafue Flats floodplain.

4.3  Thermal buffering

The problem of cold-water pollution, much like hypoxia, is driven by reservoir stratification and thus can be addressed by similar management strategies (Olden and Naiman, 2010). Most common multilevel intakes are designed so that outflows are derived from an appropriate mixture of epi- and hypolimnetic waters to meet a desirable downstream temperature threshold (Price and Meyer, 1992). A remaining challenge is a lack of understanding of what the thermal requirements are for a given river system. Rivers-Moore et al. (2013) propose a method for generating temperature thresholds for South African rivers based on time series data from dozens of monitoring stations. Unfortunately, basic monitoring data for many regions of the tropics is sparse and quite fragmentary, which further complicates the establishment of ecological thermal requirements.

4.4  Sediment manipulation

Sediment trapping by dams is not only a water quality problem, as we have discussed, but also a challenge for dam management because it causes a loss of reservoir capacity over time. Thus, in order to maintain generation capacity, many managers of hydropower dams implement sediment strategies, which include the flushing of sediments through spillways or sediment bypass systems. A recent review of sediment management practices at hydropower reservoirs provides a summary of techniques in use and evaluates their advantages and limitations, including operations and cost considerations as well as ecological impacts (Hauer et al., 2018). Sediment management can be implemented at the catchment scale, within the reservoir itself or at the dam wall. Sediment bypass systems are regarded as the most comprehensive solution, but may be expensive or infeasible because of reservoir dimensions, or cause more ecological harm than they alleviate (Graf et al., 2016; Sutherland et al., 2002).

Typically, sediment flushing is practiced in an episodic manner, creating a regime of sediment famine punctuated by intense gluts that are not so much a feast, but rather bury downstream ecosystems alive (Kondolf et al., 2014b). Environmentally optimized sediment flushes show potential for minimizing risks where these are feasible, but are rare in practice. A limitation is that not all dams are designed to allow for sediment flushes, or reservoir characteristics imposed by local geomorphology render them impractical. Furthermore, no amount of flushing is able to transport coarser bedload material (i.e., gravel or larger) downstream. To compensate for lost bedload and sediments, some managers have made the expensive effort of depositing loose gravel piles onto river margins so that they can gradually be incorporated into the downstream sediment pool as sediment-starved water inevitably cuts into banks (Kondolf et al., 2014b).

An alternative to sediment management at the site of the dam is the restoration of sediment-starved floodplain or delta wetlands, but this process is likely to be prohibitively expensive in most if not all cases. Restoration of drowning Mississippi River Delta in Louisiana, United States, are estimated to cost USD 0.5 to 1.5 billion per year for 50 years (Giosan et al., 2014).

5.1  More data from low latitudes

It is telling that, in this review focused on low latitudes, we often had to cite case studies from the temperate zone. For example, we were able to locate one study describing ecological impacts stemming from dam-driven oligotrophication at low latitudes (Granzotti et al., 2018). The simple fact is that most of the tropics and subtropics lie far from the most active research centers and there has been a corresponding gap in limnological investigations. Europe and the United States have 1.5 to 4 measurement stations for water quality per 10 000 km 2 of river basin on average. Monitoring density is 100 times smaller in Africa (UNEP, 2016). Our review found that of the 54 most-voluminous low-latitude reservoirs, 22 (41 %) have yet to be the subject of a basic limnological study to classify their mixing behavior. Further efforts to monitor river water quality and study aquatic ecology in regulated low-latitude catchments are needed to elucidate the blind spots that this review has identified.

5.2  Studies of small reservoirs

Compared with larger dams, the ecological impact of small hydropower dam systems have been poorly documented. Although small dams are likely to have smaller local impacts than large dams, the scaling of impacts is not necessarily proportional. That is, social and environmental impacts related to power generation may be greater for small dams than large reservoirs (Fencl et al., 2015). Generalizations about small hydroelectric systems are difficult because they come in so many different forms and designs. For example, nondiversion run-of-river systems will trap far less sediment than large dams, and those that do not create a deep reservoir are not subject to stratification-related effects. Thus, it is tempting to conclude that small-scale hydro will have minimal water quality impacts, but without a systematic assessment it is impossible to make a fair comparison with large-scale hydropower (Premalatha et al., 2014). Our analysis is biased towards large systems for the practical reason that larger systems are much more likely to be described in databases and studied by limnologists. Future studies on the environmental impacts of small hydropower systems would be valuable.

5.3  Better predictions of reservoir stratification behavior

Our predictions of reservoir stratification behavior based upon morphometric and hydrologic data, while helpful for understanding broad patterns of behavior, are not terribly useful for understanding water quality impacts of a specific planned dam. It would be much more useful to be able to reliably predict the depth of the thermocline, which could be compared with the depth of water intakes to assess the likelihood of discharging hypolimnetic water downstream. Existing modeling approaches to predicting mixing behavior fall into two categories: mathematically complex deterministic or process-based models and simpler statistical or semiempirical models. Deterministic models holistically simulate many aspects of lake functioning, including the capability to predict changes in water quality driven by biogeochemical processes. Researchers have used such tools to quantify impacts of reservoirs on downstream ecosystems (Kunz et al., 2013; Weber et al., 2017), but they require a large amount of in situ observational data, which is often lacking for low-latitude reservoirs. This data dependence also makes them unsuitable for simulating hypothetical reservoirs that are in a planning stage and thus they cannot inform dam environmental impact statements. A promising semiempirical approach was recently published, proposing a `generalized scaling' for predicting mixing depth based on lake length, water transparency and Monin–Obukhov length, which is a function of radiation and wind (Kirillin and Shatwell, 2016). This model was tuned for and validated against a data set consisting of mostly temperate zone lakes, so it is unclear how well it can be applied to low-latitude systems. If this or another semi-empirical model can be refined to make predictions about the stratification behavior of hypothetical reservoirs being planned, it could provide valuable information about potential risks of water quality impacts on ecosystems of future dams.

We have found that damming threatens the water quality of river systems throughout the world's lower latitudes, a fact that is not always recognized in broader critiques of large dam projects. Water quality impacts may propagate for hundreds of kilometers downstream of dams and therefore may be a cryptic source of environmental degradation, destroying ecosystem services provided to riparian communities. Unfortunately, a lack of predam data on low-latitude river chemistry and ecology makes it a challenge to objectively quantify such impacts. Building the capacity of developing countries at low latitudes to monitor water quality of their river systems should be a priority.

Seasonal stratification of low-latitude reservoirs is ubiquitous and is expected to occur in essentially any large tropical reservoir. This highlights the risk for low-latitude reservoirs to discharge cooler and anoxic hypolimnetic waters to downstream rivers depending on the depth of the thermocline relative to turbine intakes. Of course, in the absence of a randomized sampling study it is difficult to assess whether the anecdotes we have identified are outliers or part of a more general widespread pattern. Further research could investigate how common these problems are and assess the geographic or design factors that contribute toward their occurrence.

It is difficult to assess which of the water quality impacts are most damaging for two reasons. First, dams impose many impacts simultaneously and it is often difficult to disentangle which imposed water quality change is driving an ecological response, or whether multiple stressors are acting synergistically. Second, to compare the relative importance of impacts requires a calculation of value which, as we have learned from the field of ecological economics (Costanza et al., 1997), will inevitably be controversial. It does appear that water quality effects, which can render river reaches uninhabitable because of anoxia and contribute to loss of floodplain and delta wetlands through sediment trapping, exert a greater environmental impact than dam disruption to connectivity, which only directly affects migratory species.

The mitigation of water quality impacts imposed by dams has been successful in places, but its implementation is dependent on environmental regulation and associated funding mechanisms, both of which are often limited in low-latitude settings. Environmental impact assessments and follow-up monitoring should be required for all large dams. The feasibility of management actions depends upon the dam design and local geomorphology. Thus, solutions are typically custom-tailored to the context of a specific dam. We expect that as the dam boom progresses, simultaneous competing water uses will exacerbate the degradation of water quality in low-latitude river systems. Further limnological studies of data-poor regions combined with the development and validations of water quality models will greatly increase our capacity to identify and mitigate this looming water resource challenge.

All data used to produce Figs. 1, 3 and 4 are available in the ETH Zurich Research Collection ( https://doi.org/10.3929/ethz-b-000310656 ; Winton et al., 2018). The data on reservoir size and morphometry are available in the World Register of Dams database maintained by the International Commission on Large Dams, which can be accessed (for a fee) at https://www.icold-cigb.org/ (International Commission on Large Dams, 2018). The data on discharge is available in the Global Runoff Data Centre, 56068 Koblenz, Germany, accessible at https://www.bafg.de/GRDC/ (Bundesantalt für Gewässerkunde, 2018).

The supplement related to this article is available online at:  https://doi.org/10.5194/bg-16-1657-2019-supplement .

RSW, BW and EC developed the paper concept. RSW and EC extracted the data for analysis. BW provided mentoring and oversight. RSW produced the figures. RSW wrote the original draft. All authors provided critical review and revisions.

The authors declare that they have no conflict of interest.

This work is supported by the Decision Analytic Framework to explore the water-energy-food Nexus in complex transboundary water resource systems of fast developing countries (DAFNE) project, which has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement no. 690268. The authors thank Luzia Fuchs for providing graphical support for the creation of Fig. 3. Marie-Sophie Maier provided helpful feedback on figure aesthetics. Comments from two anonymous reviewers greatly improved the manuscript.

This paper was edited by Tom J. Battin and reviewed by two anonymous referees.

Anderson, E. P., Jenkins, C. N., Heilpern, S., Maldonado-ocampo, J. A., Carvajal-vallejos, F. M., Encalada, A. C., and Rivadeneira, J. F.: Fragmentation of Andes-to-Amazon connectivity by hydropower dams, Sci. Adv., 4, 1–8, https://doi.org/10.1126/sciadv.aao1642 , 2018. 

Ashley, K., Thompson, L. C., Lasenby, D. C., Mceachern, L., Smokorowski, K. E., and Sebastian, D.: Restoration of an interior lake ecosystem: The kootenay lake fertilization experiment, Water Qual. Res. J. Can., 32, 295–323, 1997. 

Bednařík, A., Blaser, M., Matoušů, A., Hekera, P., and Rulík, M.: Effect of weir impoundments on methane dynamics in a river, Sci. Total Environ., 584–585, 164–174, https://doi.org/10.1016/j.scitotenv.2017.01.163 , 2017. 

Bengtsson, L., Herschy, R. W., and Fairbridge, R. W.: Encyclopedia of lakes and reservoirs: geography, geology, hydrology and paleolimnology, Springer, 2012. 

Beutel, M. W. and Horne, A. J.: A review of the effects of hypolimnetic oxygenation on lake and reservoir water quality, Lake Reserv. Manage., 15, 285–297, https://doi.org/10.1080/07438149909354124 , 1999. 

Boehrer, B. and Schultze, M.: Stratification of lakes, Rev. Geophys., 46, RG2005, https://doi.org/10.1029/2006RG000210 , 2008. 

Bundesantalt für Gewässerkunde: Global Runoff Data Centre, available at: https://www.bafg.de/GRDC/ , last access: 1 August 2018. 

Carpenter, S. R., Caraco, N. F., Correll, D. L., Howarth, R. W., Sharpley, A. N., and Smith, V. H.: Nonpoint pollution of surface waters with phosphorus and nitrogen, Ecol. Appl., 8, 559–568, https://doi.org/10.1890/1051-0761(1998)008[0559:NPOSWW]2.0.CO;2 , 1998. 

Chícharo, L., Chícharo, M. A., and Ben-Hamadou, R.: Use of a hydrotechnical infrastructure (Alqueva Dam) to regulate planktonic assemblages in the Guadiana estuary: Basis for sustainable water and ecosystem services management, Estuar. Coast. Shelf Sci., 70, 3–18, https://doi.org/10.1016/j.ecss.2006.05.039 , 2006. 

Coble, D. W.: Fish Populations in Relation to Dissolved Oxygen in the Wisconsin River, Trans. Am. Fish. Soc., 111, 612–623, https://doi.org/10.1577/1548-8659(1982)111<612:FPIRTD>2.0.CO;2 , 1982. 

Costanza, R., d'Arge, R., de Groot, R., Farber, S., Grasso, M., Hannon, B., Limburg, K., Naeem, S., O'Neill, R. V., Paruelo, J., Raskin, R. G., Sutton, P., and van den Belt, M.: The value of the world's ecosystem services and natural capital, Nature, 387, 253–260, https://doi.org/10.1038/387253a0 , 1997. 

Coutant, C. C.: Thermal preference: when does an asset become a liability?, Environ. Biol. Fish., 18, 161–172, https://doi.org/10.1007/BF00000356 , 1987. 

Davies, B. R., Beilfuss, R. D., and Thoms, M. C.: Cahora Bassa retrospective, 1974–1997: effects of flow regulation on the lower Zambezi River, Verhandlungen des Int. Verein Limnol., 27, 2149–2157, 2001. 

Deemer, B. R., Harrison, J. A., Li, S., Beaulieu, J. J., Delsontro, T., Barros, N., Bezerra-Neto, J. F., Powers, S. M., Dos Santos, M. A., and Vonk, J. A.: Greenhouse Gas Emissions from Reservoir Water Surfaces: A New Global Synthesis Manuscript, Bioscience, 66, 949–964, https://doi.org/10.1093/biosci/biw117 , 2016. 

Delsontro, T., Kunz, M. J., Wuest, A., Wehrli, B., and Senn, D. B.: Spatial Heteogeneity of Methane Ebullition in a Large Tropical Reservoir, Environ. Sci. Technol., 45, 9866–9873, 2011. 

Deus, R., Brito, D., Mateus, M., Kenov, I., Fornaro, A., Neves, R., and Alves, C. N.: Impact evaluation of a pisciculture in the Tucuruí reservoir (Pará, Brazil) using a two-dimensional water quality model, J. Hydrol., 487, 1–12, https://doi.org/10.1016/j.jhydrol.2013.01.022 , 2013. 

Donald, D. B., Parker, B. R., Davies, J. M., and Leavitt, P. R.: Nutrient sequestration in the Lake Winnipeg watershed, J. Great Lakes Res., 41, 630–642, https://doi.org/10.1016/j.jglr.2015.03.007 , 2015. 

Eady, B. R., Rivers-Moore, N. A., and Hill, T. R.: Relationship between water temperature predictability and aquatic macroinvertebrate assemblages in two South African streams, African J. Aquat. Sci., 38, 163–174, https://doi.org/10.2989/16085914.2012.763110 , 2013. 

Fencl, J. S., Mather, M. E., Costigan, K. H., and Daniels, M. D.: How Big of an Effect Do Small Dams Have?? Using Geomorphological Footprints to Quantify Spatial Impact of Low-Head Dams and Identify Patterns of Across-Dam Variation, 10, e0141210, https://doi.org/10.1371/journal.pone.0141210 , 2015. 

Friedl, G. and Wüest, A.: Disrupting biogeochemical cycles – Consequences of damming, Aquat. Sci., 64, 55–65, 2002. 

Garnier, J., Némery, J., Billen, G., and Théry, S.: Nutrient dynamics and control of eutrophication in the Marne River system: Modelling the role of exchangeable phosphorus, J. Hydrol., 304, 397–412, https://doi.org/10.1016/j.jhydrol.2004.07.040 , 2005. 

Gibson, L., Wilman, E. N., and Laurance, W. F.: How Green is “Green” Energy?, Trends Ecol. Evol., 32, 922–935, https://doi.org/10.1016/j.tree.2017.09.007 , 2017. 

Giosan, L., Syvitski, J., Constantinescu, S., and Day, J.: Climate change: Protect the world's deltas, Nature, 516, 31–33, https://doi.org/10.1038/516031a , 2014. 

Graf, W., Leitner, P., Hanetseder, I., Ittner, L. D., Dossi, F., and Hauer, C.: Ecological degradation of a meandering river by local channelization effects: a case study in an Austrian lowland river, Hydrobiologia, 772, 145–160, https://doi.org/10.1007/s10750-016-2653-6 , 2016. 

Granzotti, R. V., Miranda, L. E., Agostinho, A. A., and Gomes, L. C.: Downstream impacts of dams: shifts in benthic invertivorous fish assemblages, Aquat. Sci., 80, 1–14, https://doi.org/10.1007/s00027-018-0579-y , 2018. 

Grizzle, J. M.: Effects of hypolimnetic discharge on fish health below a reservoir, Trans. Am. Fish. Soc., 110, 29–43, https://doi.org/10.1577/1548-8659(1981)110<29:EOHDOF>2.0.CO;2 , 1981. 

Halim, Y.: The impact of human alteration of the hydrological cycle on ocean margins, in Ocean margin processes in global change, edited by: Mantoura, R. F. C., Martin, J.-M., and Wollast, R., 301–327, John Wiley & Sons, 1991. 

Hauer, C., Wagner, B., Aigner, J., Holzapfel, P., Flödl, P., Liedermann, M., Tritthart, M., Sindelar, C., Pulg, U., Klösch, M., Haimann, M., Donnum, B. O., Stickler, M., and Habersack, H.: State of the art, shortcomings and future challenges for a sustainable sediment management in hydropower: A review, Renew. Sust. Energ. Rev., 98, 40–55, https://doi.org/10.1016/j.rser.2018.08.031 , 2018. 

Haxton, T. J. and Findlay, C. S.: Meta-analysis of the impacts of water management on aquatic communities, Can. J. Fish. Aquat. Sci., 65, 437–447, https://doi.org/10.1139/f07-175 , 2008. 

Higgins, J. M. and Brock, W. G.: Overview of Reservoir Release Improvements at 20 TVA Dams, J. Energ. Eng., 125, 1–17, https://doi.org/10.1061/(ASCE)0733-9402(1999)125:1(1) , 1999. 

Hill, L. G.: Oxygen Preference in the Spring Cavefish, Chologaster agassizi Loren, Trans. Am. Fish. Soc., 97, 448–454, https://doi.org/10.1577/1548-8659(1968)97[448:OPITSC]2.0.CO;2 , 1968. 

Holden, P. B. and Stalnaker, C. B.: Distribution and abundance of mainstream fishes of the middle and upper Colorado River basins, 1967–1973, Trans. Am. Fish. Soc., 104, 217–231, https://doi.org/10.1577/1548-8659(1975)104<217:DAAOMF>2.0.CO;2 , 1975. 

Humborg, C., Ittekkot, V., Cociasu, A., and Bodungen, B. V.: Effect of Danube River dam on Black Sea biogeochemistry and ecosystem structure, Nature, 386, 385–388, https://doi.org/10.1038/386385a0 , 1997. 

Hutchinson, G. E. and Loffler, H.: The thermal stratification of lakes, P. Natl. Acad. Sci. USA, 42, 84–86, https://doi.org/10.1073/pnas.42.2.84 , 1956. 

International Commission on Large Dams: World Register of Dams, available at: https://www.icold-cigb.org/GB/world_register/world_register_of_dams.asp , last access: 1 August 2018. 

Junk, W. J., Bayley, P. B., and Sparks, R. E.: The flood-pulse concept in river-floodplain systems, Proc. Int. Large River Symp. Can. Spec. Publ. Fish. Aquat. Sci., 106, available at: http://www.royalcommission.vic.gov.au/finaldocuments/summary/PF/VBRC_Summary_PF.pdf (last access: 18 April 2019), 1989. 

Kilham, P. and Kilham, S. S.: Endless summer: internal loading processes dominate nutrient cycling in tropical lakes, Freshwater Biol., 23, 379–389, https://doi.org/10.1111/j.1365-2427.1990.tb00280.x , 1990. 

King, J., Cambray, J. A., and Impson, N. D.: Linked effects of dam-released floods and water temperature on spawning of the Clanwilliam yellowfish Barbus capensis, Hydrobiologia, 384, 245–265, https://doi.org/10.1023/A:1003481524320 , 1998. 

Kingsford, R. T. and Thomas, R. F.: Destruction of wetlands and waterbird populations by dams and irrigation on the Murrumbidgee River in Arid Australia, Environ. Manage., 34, 383–396, https://doi.org/10.1007/s00267-004-0250-3 , 2004. 

Kirillin, G. and Shatwell, T.: Generalized scaling of seasonal thermal stratification in lakes, Earth-Sci. Rev., 161, 179–190, https://doi.org/10.1016/j.earscirev.2016.08.008 , 2016. 

Kondolf, G. M.: Hungry water: Effects of dams and gravel mining on river channels, Environ. Manage., 21, 533–551, https://doi.org/10.1007/s002679900048 , 1997. 

Kondolf, G. M., Rubin, Z. K., and Minear, J. T.: Dams on the Mekong: Cumulative sediment starvation, Water Resour. Res., 50, 5158–5169, https://doi.org/10.1002/2013WR014651 , 2014a. 

Kondolf, G. M., Gao, Y., Annandale, G. W., Morris, G. L., Jiang, E., Zhang, J., Cao, Y., Carling, P., Fu, K., Guo, Q., Hotchkiss, R., Peteuil, C., Sumi, T., Wang, H.-W., Wang, Z., Wei, Z., Wu, B., Wu, C., and Yang, C. T.: Sustainable sediment management in reservoirs and regulated rivers: Experiences from five continents, Earths Futur., 2, 256–280, https://doi.org/10.1002/2013EF000184 , 2014b. 

Kunz, M. J., Wüest, A., Wehrli, B., Landert, J., and Senn, D. B.: Impact of a large tropical reservoir on riverine transport of sediment, carbon, and nutrients to downstream wetlands, Water Resour. Res., 47, 1–16, https://doi.org/10.1029/2011WR010996 , 2011. 

Kunz, M. J., Senn, D. B., Wehrli, B., Mwelwa, E. M., and Wüest, A.: Optimizing turbine withdrawal from a tropical reservoir for improved water quality in downstream wetlands, Water Resour. Res., 49, 5570–5584, https://doi.org/10.1002/wrcr.20358 , 2013. 

Langbein, W. and Durum, W.: The aeration capacity of streams, U.S. Dept. of the Interior, Geological Survey, https://doi.org/10.3133/cir542 , 1967. 

Ledec, G. and Quintero, J. D.: Good dams and bad dams: environmental criteria for site selection of hydroelectric projects, Lat. Am. Caribb. Reg. Dev. Work. Pap. No. 16, 21, available at: http://large.stanford.edu/courses/2011/ph240/mina2/docs/Good_and_Bad_Dams_WP16.pdf (last access: 18 April 2019), 2003. 

Lewis, W. M.: A Revised Classification of Lakes Based on Mixing, Can. J. Fish. Aquat. Sci., 40, 1779–1787, https://doi.org/10.1139/f83-207 , 1983. 

Lewis, W. M.: Basis for the Protection and Management of Tropical Lakes, Lake Reserv. Manage., 5, 35–48, https://doi.org/10.1046/j.1440-1770.2000.00091.x , 2000. 

Lewis, W. M. J.: Thermal Regime Implications for Tropical Lakes, Limnol. Oceanogr., 18, 200–217, 1973. 

Lewis, W. M. J.: Tropical lakes?: how latitude makes a difference, in: Perspectives in Tropical Limnology, edited by: Schiemer, F. and Boland, K. T., 43–64, SPB Academic Publishing, Amsterdam, 1996. 

Maavara, T., Parsons, C. T., Ridenour, C., Stojanovic, S., Dürr, H. H., Powley, H. R., and Van Cappellen, P.: Global phosphorus retention by river damming, P. Natl. Acad. Sci. USA, 112, 15603–15608, https://doi.org/10.1073/pnas.1511797112 , 2015. 

Magadza, C. H. D.: Environmental state of Lake Kariba and Zambezi River Valley: Lessons learned and not learned, Lake Reserv. Manage., 15, 167–192, https://doi.org/10.1111/j.1440-1770.2010.00438.x , 2010. 

Marshall, B. E. and Junor, F. J. R.: The Decline of Salvinia-Molesta on Lake Kariba Zimbabwe, Hydrobiologia, 83, 477–484, https://doi.org/10.1007/BF02187043 , 1981. 

McCartney, M. P., Sullivan, C., and Acreman, M. C.: Ecosystem Impacts of Large Dams, International Union for Conservation of Nature and Natural Resources and United Nations Environmental Programme, 2000. 

Mumba, M. and Thompson, J. R.: Hydrological and ecological impacts of dams on the Kafue Flats floodplain system, southern Zambia, Phys. Chem. Earth, 30, 442–447, https://doi.org/10.1016/j.pce.2005.06.009 , 2005. 

Naliato, D. A. D. O., Nogueira, M. G., Perbiche-neves, G., and Unesp, P.: Discharge pulses of hydroelectric dams and their effects in the downstream limnological conditions: a case study in a large tropical river (SE Brazil), Lake Reserv. Manage., 14, 301–314, https://doi.org/10.1111/j.1440-1770.2009.00414.x , 2009. 

Nilsson, C. and Renofalt, B. M.: Linking flow regime and water quality in rivers: a challenge to adaptive catchment management, Ecol. Soc., 13, 18, 2008. 

Nurnberg, G. K.: The prediction of internal phosphorus load in lakes with anoxic hypolimnia phosphorus not considered in the mass bal- sewage diversion Of six lakes, four showed model predicted, All the outliers had anoxic phorus concentration in anoxic Lake Sam- ma, Limnol. Oceanogr., 29, 111–124, https://doi.org/10.1016/j.scr.2014.10.005 , 1984. 

Orlob, G. T.: Mathematical Modeling of Water Quality: Streams, Lakes and Reservoirs, John Wiley & Sons, Chichester, UK, available from: http://pure.iiasa.ac.at/id/eprint/2144/ (last access: 18 April 2019), 1983. 

O'Keeffe, J. H.: A Perspective on Training Methods Aimed at Building Local Capacity for the Assessment and Implementation of Environmental Flows in Rivers, Front. Environ. Sci., 6, 1–7, https://doi.org/10.3389/fenvs.2018.00125 , 2018. 

Olden, J. D. and Naiman, R. J.: Incorporating thermal regimes into environmental flows assessments: Modifying dam operations to restore freshwater ecosystem integrity, Freshw. Biol., 55, 86–107, https://doi.org/10.1111/j.1365-2427.2009.02179.x , 2010. 

Padisák, J., Barbosa, F. A. R., Borbély, G., Borics, G., Chorus, I., Espindola, E. L. G., Heinze, R., Rocha, O., Törökné, A. K. and Vasas, G.: Phytoplankton composition, biodiversity and a pilot survey of toxic cyanoprokaryotes in a large cascading reservoir system (Tietê basin, Brazil), SIL Proceedings, 1922–2010, 27, 2734–2742, https://doi.org/10.1080/03680770.1998.11898164 , 2000. 

Parker, F. L., Benedict, B. A., and Tsai, C.: Evaluation of mathematical models for temperature prediction in deep reservoirs, National Envrionmental research center, US Environmental Protection Agency, Corvallis, Oregon, 1975. 

Petts, G. E.: Water quality characteristics of regulated rivers, Prog. Phys. Geogr., 10, 492–516, https://doi.org/10.1177/030913338601000402 , 1986. 

Power, M. E., Dietrich, W. E., and Finlay, J. C.: Dams and Downstream Aquatic Biodiversity: Potential Food Web Consequences of Hydrologic and Geomorphic Change, Environ. Manage., 20, 887–895, 1996. 

Preece, R. M. and Jones, H. A.: The effect of Keepit Dam on the temperature regime of the Namoi River, Australia, River Res. Appl., 18, 397–414, https://doi.org/10.1002/rra.686 , 2002. 

Premalatha, M., Abbasi, T., and Abbasi, S. A.: Science of the Total Environment A critical view on the eco-friendliness of small hydroelectric installations, Sci. Total Environ., 481, 638–643, https://doi.org/10.1016/j.scitotenv.2013.11.047 , 2014. 

Price, R. E. and Meyer, E. B.: Water quality management for reservoirs and tailwaters: operational and structural water quality techniques, Vicksburg, Mississippi, USA, available at: https://apps.dtic.mil/dtic/tr/fulltext/u2/a254184.pdf (last access: 18 April 2019), 1992. 

Richter, B. D.: Re-thinking environmental flows: from allocations and reserves to sustainability boundaries, River Res. Appl., 30, https://doi.org/10.1002/rra.1320 , 2009. 

Rivers-Moore, N. A., Dallas, H. F., and Morris, C.: Towards setting environmental water temperature guidelines: A South African example, J. Environ. Manage., 128, 380–392, https://doi.org/10.1016/j.jenvman.2013.04.059 , 2013. 

Rolls, R. J., Growns, I. O., Khan, T. A., Wilson, G. G., Ellison, T. L., Prior, A., and Waring, C. C.: Fish recruitment in rivers with modified discharge depends on the interacting effects of flow and thermal regimes, Freshw. Biol., 58, 1804–1819, https://doi.org/10.1111/fwb.12169 , 2013. 

Rubin, J. A., Gordon, C., and Amatekpor, J. K.: Causes and consequences of mangrove deforestation in the Volta estuary, Ghana: Some recommendations for ecosystem rehabilitation, Mar. Pollut. Bull., 37, 441–449, https://doi.org/10.1016/S0025-326X(99)00073-9 , 1999. 

Sato, Y., Bazzoli, N., Rizzo, E., Boschi, M. B., and Miranda, M. O. T.: Influence of the Abaeté River on the reproductive success of the neotropical migratory teleost Prochilodus argenteus in the São Francisco River, downstream from the Três Marias Dam, southeastern Brazil, River Res. Appl., 21, 939–950, https://doi.org/10.1002/rra.859 , 2005. 

Selge, F. and Gunkel, G.: Water reservoirs: worldwide distribution, morphometric characteristics and thermal stratification processes, in: Sustainable Management of Water and Land in Semiarid Areas, edited by: Gunkel, G., Aleixo da Silva, J. A., and Sobral, M. d. C., 15–27, 2013. 

Smith, L. L., Oseid, D. M., Kimball, G. L., and El-Kandelgy, S. M.: Toxicity of Hydrogen Sulfide to Various Life History Stages of Bluegill (Lepomis macrochirus), T. Am. Fish. Soc., 105, 442–449, https://doi.org/10.1577/1548-8659(1976)105<442:TOHSTV>2.0.CO;2 , 1976. 

Smith, V. H.: Eutrophication of freshwater and coastal marine ecosystems a global problem, Environ. Sci. Pollut. Res., 10, 126–139, https://doi.org/10.1065/espr2002.12.142 , 2003. 

Spoor, W. A.: Distribution of fingerling brook trout, Salvelinus fontinalis (Mitchill), in dissolved oxygen concentration gradients, J. Fish Biol., 36, 363–373, 1990. 

Straskraba, M., Tundisi, J. G., and Duncan, A.: Comparative reservoir limnology and water quality management, Springer-Science+Business Media, 1993. 

Sutherland, A. B., Meyer, J. L., and Gardiner, E. P.: Effects of land cover on sediment regime and fish assemblage structure in four southern Appalachian streams, Freshw. Biol., 47, 1791–1805, https://doi.org/10.1046/j.1365-2427.2002.00927.x , 2002. 

Tharme, R. E.: A global perspective on environmental flow assessment: Emerging trends in the development and application of environmental flow methodologies for rivers, River Res. Appl., 19, 397–441, https://doi.org/10.1002/rra.736 , 2003. 

Thurston, R. V, Russo, R. C., and Phillips, G. R.: Acute Toxicity of Ammonia to Fathead Minnows, T. Am. Fish. Soc., 112, 705–711, https://doi.org/10.1577/1548-8659(1983)112<705:ATOATF>2.0.CO;2 , 1983. 

Todd, C. R., Ryan, T., Nicol, S. J., and Bearlin, A. R.: The impact of cold water releases on the critical period of post-spawning survival and its implications for Murray cod (Maccullochella peelii peelii): A case study of the Mitta Mitta River, southeastern Australia, River Res. Appl., 21, 1035–1052, https://doi.org/10.1002/rra.873 , 2005. 

Tolmazin, D.: Changing Coastal oceanography of the Black Sea. I: Northwestern Shelf, Prog. Oceanogr., 15, 217–276, https://doi.org/10.1016/0079-6611(85)90038-2 , 1985. 

Tundisi, J. G.: Distribuição espacial, seqência temporal e ciclo sazonal do fitoplâncton em represas: fatores limitantes e controladores, Rev. Bras. Biol., 50, 937–955, 1990. 

Turner, R. E., Qureshi, N., Rabalais, N. N., Dortch, Q., Justic, D., Shaw, R. F., and Cope, J.: Fluctuating silicate:nitrate ratios and coastal plankton food webs, P. Natl. Acad. Sci. USA, 95, 13048–13051, https://doi.org/10.1073/pnas.95.22.13048 , 1998. 

UNEP: A Snapshot of the World's Water Quality: Towards a Global Assessment, available at: http://www.unep.org/publications/ (last access: 18 April 2019), 2016. 

Van Cappellen, P. and Maavara, T.: Rivers in the Anthropocene: Global scale modifications of riverine nutrient fluxes by damming, Ecohydrol. Hydrobiol., 16, 106–111, https://doi.org/10.1016/j.ecohyd.2016.04.001 , 2016. 

Vannote, R. L. and Sweeney, B. W.: Geographic Analysis of Thermal Equilibria: A Conceptual Model for Evaluating the Effect of Natural and Modified Thermal Regimes on Aquatic Insect Communities, Am. Nat., 115, 667–695, https://doi.org/10.1086/283591 , 1980. 

Walker, K. F., Hillman, T. J., and Williams, W. D.: Effects of impoundments on rivers: an Australian case study, SIL Proceedings, 1922–2010, 20, 1695–1701, https://doi.org/10.1080/03680770.1977.11896755 , 1978. 

Ward, J. V and Stanford, J. A.: Evolutionary Ecology of Aquatic Insects, Annu. Rev. Entomol., 97–117, 1982. 

Weber, M., Rinke, K., Hipsey, M. R., and Boehrer, B.: Optimizing withdrawal from drinking water reservoirs to reduce downstream temperature pollution and reservoir hypoxia, J. Environ. Manage., 197, 96–105, https://doi.org/10.1016/j.jenvman.2017.03.020 , 2017. 

Wera, F., Ling, T., Nyanti, L., Sim, S., and Grinang, J.: Effects of opened and closed spillway operations of a large tropical hydroelectric dam on the water quality of the downstream river, Hindawi J. Chem., 2019. 

Wetzel, R. G.: Limnology: lake and river ecosystems, Academic Press, 2001. 

Winemiller, K. O., McIntyre, P. B., Castello, L., Fluet-Chouinard, E., Giarrizzo, T., Nam, S., Baird, I. G., Darwall, W., Lujan, N. K., Harrison, I., Stiassny, M. L. J., Silvano, R. A. M., Fitzgerald, D. B., Pelicice, F. M., Agostinho, A. A., Gomes, L. C., Albert, J. S., Baran, E., Petrere, M., Zarfl, C., Mulligan, M., Sullivan, J. P., Arantes, C. C., Sousa, L. M., Koning, A. A., Hoeinghaus, D. J., Sabaj, M., Lundberg, J. G., Armbruster, J., Thieme, M. L., Petry, P., Zuanon, J., Vilara, G. T., Snoeks, J., Ou, C., Rainboth, W., Pavanelli, C. S., Akama, A., van Soesbergen, A., and Saenz, L.: Balancing hydropower and biodiversity in the Amazon, Congo, and Mekong, Science, 351, 128–129, https://doi.org/10.1126/science.aac7082 , 2016. 

Winton, R. S., Calamita, E., and Wehrli, B.: Physical data for the 54 most voluminous low latitude reservoirs, https://doi.org/10.3929/ethz-b-000310656 , 2018.  

Young, W. C., Kent, D. H., and Whiteside, B. G.: The influence of a deep storage reservoir on the species diversity of benthic macroinvertebrate communities of the Guadalupe River, Texas J. Sci., 27, 213–224, 1976. 

Zarfl, C., Lumsdon, A. E., Berlekamp, J., Tydecks, L., and Tockner, K.: A global boom in hydropower dam construction, Aquat. Sci., 77, 161–170, https://doi.org/10.1007/s00027-014-0377-0 , 2014. 

Zhang, L., Li, Q.-H., Huang, G.-J., Ou, T., Li, Y., Wu, D., Zhou, Q.-L., and Gao, T.-J.: Seasonal stratification and eutrophication characteristics of a deep reservoir, Longtan Reservoir in subtropical area of China, Huan Jing Ke Xue, 36, 438–47, available at: http://www.ncbi.nlm.nih.gov/pubmed/26031068 (last access: 19 October 2018), 2015. 

Zhong, Y. and Power, G.: Environmental impacts of hydroelectric projects o n fish resources, Regulated Rivers: Research & Management, 12, 81–98, https://doi.org/10.1002/(SICI)1099-1646(199601)12:1<81::AID-RRR378>3.0.CO;2-9 , 2015. 

• Introduction
• Impacts of dams on river water quality
• How prevalent is stratification of low-latitude reservoirs?
• Managing water quality impacts of dams
• Further research needs
• Conclusions
• Data availability
• Author contributions
• Competing interests
• Acknowledgements
• Review statement

Literature Review: Real Time Water Quality Monitoring and Management

• Conference paper
• First Online: 05 January 2021
• Cite this conference paper

• Deepika Gupta 19 ,
• Ankita Nainwal 19 &
• Bhaskar Pant 19  

Part of the book series: Advances in Intelligent Systems and Computing ((AISC,volume 1254))

1321 Accesses

1 Citations

With the advent of this new era of water crisis, save water is the cry all over. Water sources are encroached from every existence on Earth. Saving water needs a systematic monitoring approach to determine its quality. Availability of Internet of Things (IoT) and remote sensing techniques mark the ease of congregating, analyzing and handling of real time data to further accelerate measures taken upon. Real-time water quality monitoring and management initiates prompt alarm ensuring timely response to water contamination in protecting and conserving the aquatic habitat, improving crop production by controlling quality of irrigated water, etc. This paper upheavals the water quality parameters required due consideration for monitoring real time water quality along with the available remote sensors. Also it briefs the review of parameters covered so far. Further it proposes the methodology suitable to the needs of detecting real time water contaminations based on the challenges of existing management system and IoT.

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Subscribe and save.

• Get 10 units per month
• Download Article/Chapter or eBook
• 1 Unit = 1 Article or 1 Chapter
• Cancel anytime
• Available as PDF
• Read on any device
• Instant download
• Own it forever
• Available as EPUB and PDF
• Compact, lightweight edition
• Dispatched in 3 to 5 business days
• Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Similar content being viewed by others

Internet of Things for Water Quality Monitoring and Assessment: A Comprehensive Review

Water Quality Monitoring and Assessment for Efficient Water Resource Management through Internet of Things and Machine Learning Approaches for Agricultural Irrigation

Smart IoT and Machine Learning-based Framework for Water Quality Assessment and Device Component Monitoring

Eleventh Five Year plan (2007–12), Planning commission government of India, vol 2

Google Scholar  

Pule M, Yahya A, Chuma J (2017) Wireless sensor networks: a survey on monitoring water Quality. ScienceDirect 15(6):562–570. https://doi.org/10.1016/j.jart.2017.07.004

Article   Google Scholar  

Radhakrishnan V, Wu W (2018) IoT Technology for smart water system. IEEE, pp 1493– 1498. https://doi.org/10.1109/hpcc/smartcity/dss.2018.00246

Robles T et al (2014) An internet of things-based model for smart water management, IEEE, pp 821–826. https://doi.org/10.1109/waina.2014.129

Cpcb.nic.in (2016) Real time water quality monitoring of river Ganga. http://52.172.40.227:8992/cr/ . Accessed 14 Jul 2014

Water Quality Data Year (2016) https://cpcb.nic.in/nwmp-data-2016/ . Accessed 28 Aug 2019

Noura M, Atiquzzaman M, Gaedke M (2018) Interoperability in internet of things: taxonomies and open challenges. Springer 24(3):796–809. https://doi.org/10.1007/s11036-018-1089-9

Sisinni E, Saifullah A, Han S, Jennehag U, Gidlund M (2018) Industrial internet of things: challenges, opportunities, and directions. IEEE X(X):1–11. https://doi.org/10.1109/TII.2018.2852491

Libelium pushes the Water Quality Market ahead with its new Smart Water Xtreme Monitoring Platform (2018). http://www.libelium.com/libelium-pushes-the-water-quality-market-ahead-with-its-new-smart-water-xtreme-monitoring-platform/ . Accessed 13 Nov 2018

Boronline®—Rolls-Royce, (2005). https://www.rolls-royce.com/~/media/Files/R/Rolls-Royce/documents/customers/nuclear/UK_Boronline . Accessed 23 Jan 2017

Chouler J, Lorenzo MD (2015) Water quality monitoring in developing countries; can microbial fuel cells be the answer? MDPI 5(3):450–470. https://doi.org/10.3390/bios5030450

JRC European Union (2013) Review of sensors to monitor water quality. https://doi.org/10.2788/35499

Download references

Author information

Authors and affiliations.

Graphic Era Deemed to Be University, Dehradun, 248002, India

Deepika Gupta, Ankita Nainwal & Bhaskar Pant

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Deepika Gupta .

Editor information

Editors and affiliations.

Department of Computer Science and Engineering, GIET University, Gunupur, Odisha, India

Raghvendra Kumar

Institute of Engineering and Technology, Thu Dau Mot University, Binh Duong, Vietnam

Nguyen Ho Quang

CMR Institute of Technology, Hyderabad, Telangana, India

Vijender Kumar Solanki

Universidad Don Bosco, Antiguo Cuscatlán, El Salvador

Manuel Cardona

KIIT Deemed to be University, Bhubaneswar, Odisha, India

Prasant Kumar Pattnaik

Rights and permissions

Reprints and permissions

Copyright information

© 2021 Springer Nature Singapore Pte Ltd.

About this paper

Cite this paper.

Gupta, D., Nainwal, A., Pant, B. (2021). Literature Review: Real Time Water Quality Monitoring and Management. In: Kumar, R., Quang, N.H., Kumar Solanki, V., Cardona, M., Pattnaik, P.K. (eds) Research in Intelligent and Computing in Engineering. Advances in Intelligent Systems and Computing, vol 1254. Springer, Singapore. https://doi.org/10.1007/978-981-15-7527-3_88

Download citation

DOI : https://doi.org/10.1007/978-981-15-7527-3_88

Published : 05 January 2021

Publisher Name : Springer, Singapore

Print ISBN : 978-981-15-7526-6

Online ISBN : 978-981-15-7527-3

eBook Packages : Intelligent Technologies and Robotics Intelligent Technologies and Robotics (R0)

Share this paper

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Publish with us

Policies and ethics

• Find a journal
• Track your research