Assessment of water quality of the Narmada River (India) using the fuzzy water quality index method

Narmada River is the lifeline of Madhya Pradesh and it is the fifth-longest river in the Indian subcontinent. River water is the main resource of fresh water, crucial for the survival of civilizations, as it provides reliable water for our domestic usage and various activities in agriculture, transportation, and industrial processes. Due to their socio-economic and ecological importance, freshwater resources lay the foundation for the growth and progress of the nation (Wang & Yang 2019). As the global population increases, the demands on the world's freshwater supplies continue to grow. Hence, effective and sustainable management of water resources is vital for ensuring overall sustainable development. Along the banks of the Narmada River, numerous small towns and major cities have emerged. Unfortunately, these human settlements have inadvertently contributed to water pollution through the discharge of untreated and partially treated domestic and industrial waste, as well as natural drainage and the urban river system receiving various types of contaminations including organic pollutants, carcinogenic substances, heavy metal, pathogens, and pharmaceutical drugs which cause serious health issue about human and environment.

In developing nations, determining the quality of river water has become a crucial concern in recent years. It is common knowledge that ensuring the availability of clean and safe drinking water is essential for the well-being of both human populations and the environment (Swain & Sahoo 2017). Pure water resources around the world are under threat from anthropogenic and industrial activity's growing contamination. Thus, comprehensive and accurate assessments of trends in water quality as well as source identification of pollutants have emerged as the need of the hour.

Monitoring of surface water sources is essential for generating reliable information on its water quality which in turn will go a long way in preventing and controlling its pollution. Consistent monitoring data becomes highly essential for such assessments and evolving effective monitoring strategies. A major problem in water quality monitoring is the handling of huge and complex data sets generated due to the large number of water quality variables at different monitoring stations. Therefore, there has been an increase in the need for ongoing river water quality monitoring in concern (Sharma et al. 2017; Swain & Sahoo 2017).

However, the evaluation of water quality samples is difficult, because each sample has many water quality parameters. To resolve this problem, Horton (1965) proposed the first water quality index (WQI) by weighting some water quality variables. The WQI is a valuable and unique rating to represent the overall water quality status in a single value which provides its suitability of water for various purposes such as drinking, irrigation, and industrialization, and water quality status in simple terms such as excellent, good, poor, very poor, and unsuitable for drinking.

Studies of WQI have been reported on many rivers, e.g., Ganga River (Bhargava 1983); Dez River (Samaneh et al. 2020); River Mahi (Srivastava et al. 2011); River Cauvery (Pandian et al. 2011); River Sabarmati (Shah & Joshi 2017); Al-Gharraf River (Ewaid & Abed 2017); and Yamuna River (Sharma & Kansal 2011).

The WQI has also been applied for surface water quality assessment all around the world in the last few decades (Nikoo et al. 2011; Pandian et al. 2011; Srivastava et al. 2011; Kumar Meher et al. 2015; Bora & Goswami 2017; Ewaid & Abed 2017; Shah & Joshi 2017; Chacón et al. 2018).

There is a need for more advanced methods to assess the water quality variables. In this context, a new, alternative method is being developed by using artificial intelligence techniques. The use of artificial intelligence techniques has boosted the demand for water quality modeling to circumvent this approach (Chau 2006). Fuzzy logic is a noteworthy example of artificial intelligence techniques that are used to enhance and optimize human experience (Hameed et al. 2017; Yaseen et al. 2018). The concept of fuzzy logic was initially introduced by Professor Zadey in 1965 and has since found widespread application across various domains. According to Zadeh, fuzzy logic proves to be highly suitable for developing environmental indices due to its capacity to handle nonlinear, uncertain, ambiguous, and subjective data. The fuzzy water quality index (FWQI) method represents a significant advancement in water quality assessment, leveraging the principles of fuzzy logic to offer a more nuanced and accurate evaluation of water quality status. Hence, FWQI is an effective model that not only provides the index value but also establishes the relationship between the input and output parameters by using fuzzy inference rules.

In the traditional method, WQI calculations are necessary, whereas in the FWQI method, a model is developed, values are inputted, and the WQI is determined. Therefore, comparing FWQI and WQI reveals that the FWQI method is a reliable, affordable, effective, and alternative method for the evaluation of water quality analysis (Oladipo et al. 2021).

Numerous research using the FWQI approach have been published during the past few decades: FWQI in Ribeira do Iguape River (Lermontov et al. 2009); WQI development using fuzzy logic in Karoon River (Babaei Semirom et al. 2011); water quality management (Che Osmi et al. 2016); water quality classification (Dewanti & Abadi 2019); river-pollution decision support expert system (Nasiri et al. 2007); fuzzy water pollution index in Qu River (Li et al. 2016); application of adaptive neuro-fuzzy to estimate BOD of Surma River (Ahmed & Shah 2017); fuzzy logic inference index in Tigris River (Ewaid et al. 2019); and WQI using fuzzy logic in Utcubamba River (Quiñones-Huatangari et al. 2020).

Narmada River is an important river in India and this was the first comprehensive water quality modeling study on the Narmada River. The study was finished using routine sampling over 5 years at six sample stations. Consequently, FWQI has emerged as a powerful and versatile approach for conducting a more holistic evaluation of water quality.

The Narmada River, the largest westward-flowing river in India, originated from Amarkantak, a small region in Madhya Pradesh characterized. It spans a total length of 1,312 km, with the initial 1,077 km (constituting 86.17%) flowing through in Madhya Pradesh, followed by the final 161 km (10.6%) in Gujrat. The remainder of its course extends through Maharashtra (1.5%) and Chhattisgarh (0.72%). This river serves as a vital resource for various purposes, including domestic, irrigation, and industrial applications.

Water samples were gathered over 5 years spanning from April 2017 to March 2018, April 2018 to March 2019, April 2019 to March 20120, April 2020 to March 2021, and April 2021 to March 2022.

Samples were processed using standard techniques for the preparation, preservation, and analysis of water and wastewater, as outlined by the American Public Health Association in 2005 (APHA 2005). A comprehensive breakdown of parameter units, permissible range, and methods of measurement is presented in Table 1.

In this context, the parameters a, b, c, and d are used, where a and d denote the base while b and c represent the height of the trapezoidal.

The membership function was divided into five linguistic categories: Excellent, Good, Poor, Very Poor, and Unsuitable for drinking. These categories were designed to ensure uniformity according to the rating scale for all variables. Triangular membership functions were employed for intervals II, III, and IV, whereas trapezoidal membership functions were used for intervals I and V.

The detailed description of all input and output linguistic variables with their ranges and intervals is represented in Table 2.

A fuzzy rule base is composed of if-then rules that combine input and output variables. Rules and their quantities are established through the utilization of expert knowledge, personal experience, and optimization of outcomes (Pratihar et al. 1999; Lermontov et al. 2009; Li et al. 2016; Oladipo et al. 2021). Twenty-five rules were applied in this study, and detailed descriptions of five rules are as follows:

Rule 1: If the following conditions are met: excellent turbidity; excellent pH; excellent DO; excellent BOD₅; excellent TDS; excellent TSS; excellent COD; excellent EC; good TH; excellent TA; and excellent chloride; then the FWQI is excellent.

Rule 2: If the following conditions are met: good turbidity; good pH; excellent DO; excellent BOD₅; excellent TDS; excellent TSS; excellent COD; good EC; excellent TH; excellent TA; and excellent chloride; then the FWQI is good.

Rule 3: If the following conditions are met: good turbidity; good pH; good DO; good BOD₅; good TDS; good TSS; good COD; good EC; good TH; good TA; and good chloride; then the FWQI is good.

Rule 4: If the following conditions are met: good turbidity; good pH; good DO; good BOD₅; excellent TDS; excellent TSS; excellent COD; good EC; excellent TH; excellent TA; and excellent chloride; then the FWQI is good.

Rule 5: If the following conditions are met: good turbidity; good pH; good DO; good BOD₅; good TDS; excellent TSS; excellent COD; good EC; good TH; excellent TA; and good chloride; then the FWQI is good.

In the third step, output was obtained by using defuzzification which was based on the center of gravity method (Oladipo et al. 2021). The ‘fuzzy logic toolbox’ in MATLAB 2015 has been used to process each of these computations.

Turbidity measures liquid clarity; higher turbidity means more cloudiness due to suspended or dissolved particles. High turbidity in drinking water is visually unappealing. Turbidity is measured in Nephelometric Turbidity Units (NTUs) using a turbidimeter, the acceptable level for drinking water is 5 NTU. The average turbidity concentrations ranged from 6.59 to 7.49 NTU, peaking at station third in 2019–2020 and reaching their lowest at station fifth in 2020–2021.

pH measures a solution's acidity or alkalinity based on hydrogen ion (H⁺) concentration. The pH scale ranges from 0 to 14, with 7 being neutral. Values below 7 indicate acidity, and values above 7 indicate alkalinity. Drinking water should typically have a pH between 6.5 and 8.5. The average pH concentration fluctuated from 7.16 to 7.95, with the highest recorded at station second in 2020–2021 and the lowest at station six in 2017–2018.

Dissolved oxygen (DO) measures the amount of oxygen dissolved in water, crucial for aquatic organisms' survival. Ideally, DO levels in freshwater ecosystems should surpass 5 mg/l to support most aquatic life. The average concentration of DO varies from 7.73 to 8.93 mg/l, with the highest concentration at station second in 2020–2021 and the lowest concentration at station fifth in 2017–2018.

Biochemical oxygen demand (BOD₅) evaluates the oxygen required by aerobic organisms in water to decompose organic matter within a set time and temperature. Elevated BOD₅ levels indicate higher organic pollution. The average concentration of BOD₅ varies from 1.23 to 2.01 mg/l, with the highest concentration at station third in 2017–2018 and the lowest at station fifth in 2020–2021.

Total dissolved solids (TDS) in water reveal the presence of minerals, metals, organic matter, and salts, encompassing ions such as magnesium, calcium, sodium, and potassium. TDS quantifies the overall concentration of dissolved solids, typically measured in mg/l. The accepted limit for TDS in drinking water is 500 mg/l. The average concentration of TDS exhibited a range from 142 to 179 mg/l, reaching its highest at station third in 2017–2018 and its lowest at station fifth in 2020–2021.

Total suspended solids (TSS) in water include suspended solid particles, both organic and inorganic, such as sediment, silt, and plankton. TSS is pivotal for water quality assessment due to its adverse effects on aquatic ecosystems such as reduced light penetration and habitat suffocation. The accepted limit for TSS in drinking water is 75 mg/l. The average concentration of TSS spanned from 27 to 44 mg/l, with the greatest concentration at station third in 2017–2018 and the lowest at station fifth in 2021–2022.

Chemical oxygen demand (COD) measures the oxygen equivalent of organic matter in a water sample that can be oxidized by a strong chemical oxidant. The accepted limit for COD in drinking water is typically 200 mg/l. The average concentration of COD varied from 15.78 to 22.96 mg/l, with the highest at station third in 2017–2018 and the lowest at station fifth in 2020–2021.

Electrical conductivity (EC) measures a solution's capacity to conduct electrical current, reflecting the presence of dissolved ions such as salts, minerals, and other conductive substances in water quality. Elevated EC levels typically signify increased dissolved solids in the water. EC value should not exceed 300 μs/cm to ensure good drinking water. The average concentration of EC varies from 271 to 309 μs/cm, reaching its peak at station three in 2021–2022 and its lowest at station two in 2020–2021.

Total hardness (TH) is the sum of the calcium and magnesium concentrations, both expressed as calcium carbonate, in mg/l. The accepted limit for TH in drinking water is typically 300 mg/l. The average concentration of TH varies from 96 to 127 mg/l, with the highest recorded at station three in 2017–2018 and the lowest at station five in 2020–2021.

Total alkalinity (TA) measures water's ability to neutralize acids and maintain a stable pH, thereby resisting acidification. The accepted limit for TA in drinking water is typically 200 mg/l. The average concentration of TA ranged from 19 to 38 mg/l with the greatest concentration at station third in 2017–2018 and the lowest at station second in 2020–2021.

Chlorides are soluble mineral compounds dissolved by water as it filters through the earth. Public drinking water standards require that chloride levels do not exceed 250 mg/l. The average concentration of chloride varied from 11.92 to 16.63 mg/l, reaching its highest at station third in 2017–2018 and its lowest at station second in 2020–2021.

The concentration of turbidity increased during the monsoon season due to nearby agricultural runoff, impacting its average concentration. Moreover, increased urban activity contributed to elevated concentrations of BOD₅, TDS, TSS, COD, TH, TA, and chloride. Table 3 presents the results of the average concentration of parameters at sampling stations spanning from 2017 to 2022.

Unit weights are allocated based on the National Sanitation Foundation Water Quality Index (Brown et al. 1970). The unit weight values for turbidity, pH, DO, BOD₅, TDS, TSS, COD, EC, TH, TA, and chloride are 0.08, 0.11, 0.17, 0.11, 0.06, 0.01, 0.11, 0.07, 0.07, 0.11, and 0.09, respectively.

Using MATLAB, the FWQI values were computed for various stations for the period from 2017 to 2022. In the years 2017–2018, the value of FWQI at stations first, second, third, fourth, fifth, and sixth were 62.50, 59.60, 62.50, 62.50, 50.00, and 50.00, while the WQI values at the same stations were 59.66, 58.98, 60.97, 62.51, 47.17, and 47.52. According to both methods, the first four stations were classified as having poor water quality, whereas the last two stations were categorized as having good water quality during that period.

In the years 2018–2019, the FWQI values at the same stations were 61.00, 58.10, 62.30, 60.40, 50.00, and 49.90, and the WQI values were 58.79, 57.98, 59.62, 60.05, 49.67, and 48.82. Once again, the first four stations were considered to have poor water quality, while the last two stations had good water quality.

In the years 2020–2021, the FWQI values at stations first, second, third, fourth, fifth, and sixth were 60.00, 54.20, 61.90, 62.30, 50.00, and 50.00, with WQI values of 58.19, 58.16, 59.38, 57.53, 47.49, and 50.13. According to both methods, the first four stations were considered to have poor water quality, while the last two stations had good water quality.

Lastly, in the years 2021–2022, the FWQI values at the same stations were 60.40, 57.60, 61.70, 61.80, 50.00, and 49.90, while the WQI values were 58.54, 58.21, 59.89, 60.19, 49.03, and 49.10. Similar to previous years, the first four stations were considered to have poor water quality, while the last two stations had good water quality.

The maximum FWQI value of 62.50 occurred at stations first, third, and fourth in 2017–2018 and 2019–2020, while the minimum value of 49.90 was recorded at station six in 2018–2019 and 2021–2022. The highest WQI value of 62.51 was observed at station four in 2017–2018, and the lowest value of 46.52 was noted at station six during the same period.

The main objective of this investigation was to assess the river water quality by developing a model using the FWQI method at six designated sampling stations. This model incorporated 11 input parameters (turbidity, pH, DO, BOD₅, TDS, TSS, COD, EC, TH, TA, and chloride), a single output (FWQI), a fuzzy inference rule, and the center of gravity approach for defuzzification. Except for turbidity and EC, the concentrations of all parameters were within permissible limits. Turbidity levels exceeded the acceptable limit during the monsoon season, resulting in elevated concentrations. The value of EC was just over the acceptable range at station third during the periods of 2017–2018 and 2021–2022. According to the average FWQI results, the order of contamination was S-3 > S-4 > S-1 > S-2 > S-5 > S-6. Based on the average WQI values, the contamination order was S-3 > S-4 > S-1 > S-2 > S-6 > S-5. Both methods indicated that the third station was significantly affected by urban activities, while the fifth and sixth stations were minimally impacted. Water quality at the first to fourth stations fell into the poor category, whereas the fifth and sixth stations were categorized as having good water quality. This study represents the first water quality modeling investigation of the Narmada River and provides valuable data for future research initiatives. Consequently, FWQI emerges as a reliable, expeditious, and effective modeling tool for river basin management action plans.

The author is grateful for the support given by the Principal, Jabalpur Engineering College, Jabalpur, Madhya Pradesh for the support given for conducting the study.

Conceptualization, acquisition of data, preparation of manuscript, graphs, and tables, software: D. C. R.; Analysis and interpretation of data, review and editing, visualization: R. C.; Drafting and analysis, resources, validation, review, and manuscript preparation: A. V.; and final approval of manuscript by all authors.

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Data may be shared upon request.

The authors declare there is no conflict.