Suitability of drinking water quality in Chittagong Metropolitan City, Bangladesh: research on urban water bodies (UWBs) using multivariate analytic techniques

Fresh water is an essential natural resource; it provides immense socioeconomic and ecological benefits for all forms of life on earth, and plays a vital role in maintaining the health and welfare of human beings (WHO & UNCF 2000; Agnew & Woodhouse 2011; Muhangane et al. 2017). However, this precious resource is scarce in nature, contains only 0.01% of the total hydrosphere, and water stored in small urban water bodies (UWBs)¹ is only a tiny fraction of it. Moreover, around 780 million people in the world do not have access to clean and safe water. As a consequence, around 6–8 million people die each year due to waterborne diseases and disasters. Furthermore, water pollution is one of the major environmental concerns of our time (Carr & Neary 2008). The UWBs are susceptible to various pollutants depending on physical processes and anthropogenic activities. The problem is serious for drinking water; public health is at high risk due to the presence of chemical contaminants. Prolonged exposure to contaminated water has been known to increase the risks of diarrheal diseases, cancer, and disorders in the kidney, liver, and reproductive organs (WHO 2002; Fawell & Nieuwenhuijsen 2003). The concern over the ensuing freshwater supply has compelled developing countries to search for alternative water sources. The UWBs can play a vital role in this regard as alternative sources of water for household consumption (United Nations 2010; Rahman et al. 2011). Ensuring the safety of drinking water is, therefore, a growing public concern. Drinking water is now recognized as a fundamental human right by the United Nations. Availability and sustainable management of good-quality water were set as one (Number 6) of the UN Sustainable Development Goals (SDGs) (United Nations 2015). An understanding of water quality² and its availability is vital because waterborne diseases are still a major cause of death in many parts of the developing world (WHO 2002, 2011). Therefore, monitoring and control of drinking water quality is a top-priority policy agenda in many parts of the world (UNESCO 2013).

Chittagong is one of the major developing metropolises, an important harbor and seaport city, and Bangladesh's manufacturing and commercial heart. There are a lot of large polluting industries in the city (DoE 2004). The city has recently seen the effects of increasing urbanization. Over the previous three decades, the overall quantity of UWBs in metropolitan has steadily decreased at a degree of 10% each year. In the previous 30 years, over 56% of the land cover had changed, owing mostly to the increase in developed areas and other anthropogenic activities (Molla et al. 2020). As a result, concerns about potable water quality and paucity apply to the water itself and the number of threats associated with harmful material diffusion into freshwater environments. In reality, access to clean water has become a critical concern for the city dwellers in Chittagong Metropolitan City (CMC). The Chittagong Water Supply and Sewerage Authority (CWASA) is the sole organization that supplies water to city dwellers through its limited distribution networks. However, the organization can only produce 450 core liters, or 4,500 MLD, of water each day; around 80% of the water is produced from surface water sources, and the remaining 20% is produced from underground water sources. Only 3.1 million individuals had access to piped water through their house connection, and about 2.03 million people were unable to access CWASA sources of drinking water last year (CWASA Annual Report 2020–2021). Many city residents still experience a severe water shortage and obtain their water from private sources or other sources, such as natural reservoirs like ponds, canals, and rainwater catchments. Due to ineffective management techniques, city people have been experiencing erratic, insufficient, and hazardous water delivery in these conditions (De 2020). Day by day, things are becoming worse, especially in CMC's low-income residential areas (Molla et al. 2014; Molla & Chowdhury 2021).

This paper was prepared as part of a recent comprehensive field inquiry entitled ‘Management of Urban Water Bodies (UWBs) in Bangladesh: A Case Study of Chittagong Metropolitan City.’ Table 1 shows the seasonal change of eleven (11) physiochemical parameters assessed throughout the rainy and winter seasons. The observed mean values of the examined parameters reveal a significant seasonal fluctuation between the two seasons: monsoon and winter. Except for nitrate, the mean values of 10 of the 11 parameters were found to be greater in the rainy season than in the winter (Table 1). This served as a foundation for deciding on a season for additional investigation, namely a multivariate study of water quality characteristics during the monsoon season.

Usually, drinking water quality refers to the characteristics of water, whether good or bad; it shows the suitability of water for human consumption; it is frequently used in a comparative sense, and its meaning is relative. In order to conduct an interpretive analysis, credible reference points known as ‘standards’ are required; they are considered authoritative assessments of risk exposure (WHO 1995; Kannan 1997). Some reputed recommending authorities have provided information on drinking water standards and unit weights in Table 2. If values surpass permissible limits, human health is jeopardized (Carr & Neary 2008). The World Health Organization (WHO), Bureau of Indian Standards (BIS 2012), Indian Council for Medical Research (ICMR), and Bangladesh Standard Testing Institute are among the recommending agencies used as references in this study (BSTI).

There is an abundance of literature on water quality monitoring and assessment. Of the multivariate methods, principal component analysis (PCA) and cluster analysis (CA) have been successfully used by many researchers as the more trustworthy technique for data reduction (Astel et al. 2007, 2008; Simeonova & Simeonov 2007). A number of academicians and researchers (especially Singh et al. 2004; Zhang et al. 2009; Prasanna et al. 2012; Bhuiyan et al. 2016; Howladar et al. 2017; Rahman et al. 2017; Ahmad et al. 2020; Akhtar et al. 2021) have conducted research on water quality applying multivariate techniques including PCA and CA at home and abroad. However, until recently, none of them has applied these methodologies to UWBs in CMC. Such an investigation is intended to aid in raising awareness at all levels (public and private), particularly among city inhabitants, about the importance of investigating alternative water sources for home consumption and public health protection. Based on the preceding concept, our objective for the study is three-fold as follows: (i) Collection of samples for drinking water from selected sample stations (UWBs: N = 21) in CMC, (ii) determination of drinking water quality by suitability analysis (SA) using a set of physiochemical and biological parameters including certain trace metals for the rainy season; and (iii) application of multivariate techniques for data reduction, i.e. PCA and CA.

To detect existing UWBs on the Chittagong City Corporation (CCC) Map, a time series of Landsat data (Landsat™ images 30 m spatial resolution) collected from Bangladesh's Space Research and Remote Sensing Organization (SPARRSO) was used. The actual number of UWBs had also been identified using the ‘Ground Truthing Method’ (1,249). In the research region, a total of 21 sample sites (UWBs) were chosen. The sampling stations, ward number, spatial location, area in square feet, and pollution status for each site are listed in Supplementary Appendix I. The locations of the surface water sample sites are shown in Figure 1 on the CCC map. The sampling stations were selected with the spot technique. Moreover, the absolute position of the sample location was determined using the geographical positioning system (GPS). Furthermore, the sample size was identified using a number of criteria, including the sample sites' (UWBs) connectedness with city residents, their proximity to recreational and other facilities, their potential to retain water permanently, their frequency of use (which includes ablutions, bathing, dressing, cleaning, dishwashing, somewhat drinking, and fishing), and their proximity to water bodies in lower-income residential areas. These are a few of the important factors that were taken into consideration when selecting sample stations from the entire UWB distribution map (Supplementary Appendix I).

Water quality is determined by a number of factors, including physical, chemical, microbiological, and trace metals, all of which may be quantified in terms of their appropriateness for a certain function. The relative importance of one set of these factors over another would be determined mostly by the purpose for which they are to be used. The list of twenty-three (23) water quality metrics used to determine surface water quality in CMC is presented in Table 3. The table also provides the study's measuring units and analytical methodologies.

In order to obtain samples from the sampling stations (N = 21 UWBs), 23 parameters were used (Table 3); each sample, which was collected in a 200 ml polyethylene bottle, was obtained at a depth below the water's surface. The bottles were cleansed with water to be examined before collecting the samples, and the collected samples were stored by acidifying pH 2 with HNO₃ and stored at 40 °C until analysis was done. The samples were subsequently tested at the Bangladesh Council of Scientific and Industrial Research (BCSIR Laboratory) in Chittagong's Chemistry Laboratory, Industrial Microbiological Research Division.

The acceptability of drinking water in CMC UWBs was tested using a set of fifteen (15) physiochemical and microbiological parameters. With the help of SPSS, these parameters were tested and evaluated using certain common scientific and statistical procedures. For the rainy season, descriptive statistics such as observed mean value, units of measurement, standard deviation, range-maximum and minimum, suitability measures, and many more are included in the analyzed data.

A correlation analysis was performed for fourteen (14) water quality parameters, obtained from 21 spatial observation units/sample stations – UWBs in CMC, as a piece of necessary background information for multivariate analysis. The Pearson's correlation coefficient matrix (14 × 21 = 294) for the analyzed parameters is presented in Table 6. The matrix was utilized to recognize the relationship between sample parameters. The correlation coefficient matrix, which serves as a context and rationale for multivariate analysis, quantifies how well each parameter's variance explains.

As a way to unravel the primary structure in the data matrix two contrasting, but complementary multivariate analytical techniques – namely trend seeking ordination, i.e. PCA and recognition of group structure, i.e. CA were employed in the dataset. The data were analyzed using the IBM Statistical Package for the Social Sciences (SPSS), Statistics Version 25 for Windows, and CAP version: 5.0.0.465. A detailed treatment of multivariate methods such as PCA and CA may be found in Orlóci (1978), Howladar et al. (2017), and Costa et al. (2020). Prior to subjecting the data to the aforementioned analytical methods, the raw data was centered and standardized using standard score transformation to have 0 mean and unit variance.

The PCA is an extremely effective approach used to reduce the dimensionality of a dataset made up of many interconnected variables. PCA makes an effort to explain the correlations between the observations in terms of factors that are not immediately evident (Yu et al. 2003). In order to reduce the amount of data, the original dataset is converted into a new set of variables known as the principal components (PCs). These components, which are ranked in decreasing order of significance, are weighted linear combinations of the original variables that are orthogonal (non-correlated) to one another (Wunderlin et al. 2001). Typically, the first three PCs often explain a significant amount of the total variance. The Varimax rotated PCA matrix solution was sought using the SPSS package. The r-mode technique was performed using covariance as the resemblance function, which minimized the number of variables (parameters in this study) with high loading on each component, thus facilitating the interpretation of PCA results (Gotelli & Ellison 2004). Details of the technique may be found in a number of publications including Orlóci (1978), Fewster & Orloci (1978), Gotelli & Ellison (2004), Howitt & Cramer (2005), Dragovıć et al. (2008), Franco-Uría et al. (2009), Bhuiyan et al. (2016), Billa et al. (2016), Vega et al. (1998), and Helena et al. (2000).

The next step of the data reduction technique is CA. The fundamental goal of clustering is to arrange a collection of items (such as sample stations or UWBs, the spatial observation units) so that objects in the same group, or cluster, are more similar to one another than units in other categories (clusters). To put it another way, the generated object clusters should have strong external heterogeneity (across groups) and high internal heterogeneity (within the group) (McGarial et al. 2000). In this way alone, the CA is a powerful data-mining tool for researchers to identify discrete groups. Hierarchical agglomerative clustering is the method that is used the most frequently in cluster analysis. The outcomes of the CA assist with interpreting the data and point out patterns (Vega et al. 1998).

Using the Ward-algorithmic approach of clustering, the dataset's built-in group structure was revealed. This is an algorithm – a polythetic agglomerative method of clustering, which hierarchically groups the cases so that those with greater similarity (lesser distance) cluster together earlier in the numeric manipulation of the data. For calculating the distance between clusters of similar characteristics, the Euclidean rescaled linkage distance was used as the resemblance function (a measure of similarity; Singh et al. 2004; Howladar et al. 2017; Costa et al. 2020). We used a fantastic idea called a Dendrogram, which is a graphical representation of the cluster solution, to determine the number of clusters for hierarchical clustering. More specifically, a Dendrogram is a tree-like diagram that illustrates the distance between groups when they are linked and records the sequences of merges and splits. The stages of hierarchical clustering are easily recognizable just from the Dendrogram. However, determining the number of clusters by looking at the classification tree is a subjective process. Generally, one begins by looking for ‘gaps’ between joining along the horizontal axis. A Dendrogram can be understood by concentrating on the height at which any two things are connected to one another.

The results of the suitability research are presented and summarized in Table 4 for fifteen (15) physiochemical and microbiological criteria.

The observed mean value of the parameters of interest was compared to their assigned standard value to assess the water quality. The mean value indicates that the maximum sample size has exceeded the allowable limit. As shown in Table 4, 8 of the 15 water quality criteria were determined to be unfit for drinking during the rainy season. The following parameters are all above the permitted limit: temperature, EC, BOD, COD, Turbidity, Nitrate, and two biological metrics, Total and Fecal coliform.

Table 5 sums up and illustrates the findings of the suitability study for detecting trace metals in CMC's ample stations (UWBs). For the rainy season, a total of eight (8) factors were used. Except for one parameter, lead (Pb), which has levels over the permitted limit, the other trace element content of UWBs is not concerning for drinking purposes. The examined water bodies had no copper (Cu) or cadmium (Cd), however, arsenic (As) levels were found to be below detected levels (BDLs). The samples also contained chromium (Cr), iron (Fe), mercury (Hg), and manganese (Mn). However, it has been demonstrated that drinking is safe during the rainy season.

Table 6 displays Pearson's correlation matrix. The observed parameters have a substantial relationship. For instance, EC has a significant correlation with TDS (r = 0.964), TSS (r = 0.572), BOD (r = 0.661), Chloride (r = 0.565), and Salinity (r = 0.580). Then, BOD has a strong correlation with EC (r = 0.661), TDS (r = 0.681), TSS (r = 0.489), COD (r = 0.636), and Nitrite (r = 0.595). After that TDS (r = 0.487), TSS (r = 0.476), and BOD (r = 636) all have a moderate correlation with CODs. Similarly, Nitrite has a strong correlation with BOD (r = 0.595) and Pb (r = 0.449). In addition, Chloride has a strong relationship with EC (r = 0.565), Salinity (r = 0.998), and Pb (r = 0.473). Also, EC (r = 0.580), TDS (r = 0.435), Chloride (r = 0.998), and Pd (r = 0.465) all have a moderate correlation with salinity. Besides, TDS is also linked to EC (r = 0.964), TSS (r = 0.601), BOD (r = 0.681), COD (r = 0.487), and salinity (r = 0.435). Moreover, TSS also has a strong association with EC (r = 0.572), TDS (r = 0.601), BOD (r = 0.489), and COD (r = 0.476). Additionally, DO has a strong correlation with pH (r = 0.449). Furthermore, pH has a modest correlation with DO (r = 0.449). As well, Pb, like Chloride (r = 0.473), Salinity (r = 0.465), and Nitrite (r = 0.449), has a modest correlation. The results of this discovery served as a foundation for multivariate analysis, such as PCA and CA.

The PCA approach was used to detect potential pollutants sources in the sample locations. By identifying a new collection of variables known as principal components, the PCA allows for an objective summary of the water quality metrics (Factors). The consecutive components are orthogonal to one another, and each component is a linear grouping of the initial variables (parameters). Table 7 summarizes the findings of Varimax rotated PCA during the rainy season. As shown in the table, the analysis included fourteen (14) physiochemical water quality indicators collected from CMC's 21 sample stations (UBS). The raw data were normalized with standard score transformations to have a 0 mean and unit variance before being subjected to PCA. The table also includes the eigenvalue, percentage of variation, and cumulative percentage of all the variables (14 parameters) loaded on the first four (4) main components (i.e. component loadings).

The correlation matrix suggests that the original dataset may have some key underlying dimensions, referred to as Factors. By extracting new variables termed Factors-dimensions from the implicit factors in samples (extracted by PCA), the original dataset was reduced. The eigenvalues of four factors were retrieved from UWB's dataset. Table 7 shows the calculated factor loadings, as well as the percentages of variation explicated by each component and the cumulative percentage. Factor I (PC1), Factor II (PC2), Factor III (PC3), and Factor IV (PC4), respectively, explain roughly 25, 19, 15, and 14% of the overall variation. In cumulative terms, the first four (4) principal components (PC1–4), which may also be utilized as Factors (I–IV), explain 73.1% of the overall variation in the dataset (Hanssen et al. 1980; Steinnes & Henriksen 1993; Prasanna et al. 2010).

With an eigenvalue of 4.45, it accounts for 25% of the overall variance in the dataset; it is predominantly a consequence of five parameters: TDS (0.89), TSS (0.85), EC (0.85), BOD (0.69), and COD (0.62), all of which have strong positive component loadings (coefficient of association). The presence of (TDS) inorganic salts and tiny quantities of organic matter in the solution is described by this factor, whereas TSS refers to particles bigger than 2 microns present in the water column. The EC is a metric for determining what is dissolved in water. Because the existence of dissolved salts like nitrate, sulfate, and other inorganic compounds affects water, electrical conductivity rises as salinity rises. TDS and TSS have large loadings on PC1, but they do not exceed the maximum permissible concentration, hence they are safe. EC, BOD, and COD, on the other hand, were found to be unfit for consumption. When biological organisms degrade organic substances in water, they use a certain amount of dissolved oxygen (DO). A higher BOD means that more oxygen is required, which results in less oxygen being available for organisms to consume and worsen water quality. When a water sample is chemically oxidized, COD is the quantity of oxygen consumed. Higher COD values indicate that the sample contains more oxidized organic material.

With an eigenvalue of 2.32, it explains around 19% of overall variance and is dominated by two parameters: chloride (0.92) and salinity (0.92), as seen by their significant positive component loadings on PC2. Dissolved solids contain a lot of chlorides. Chloride concentrations in UWBs that are too high can be hazardous to aquatic life. Drinking water quality and ecosystem health can be harmed by high levels of dissolved salt in water.

With an eigenvalue of 2.13, it accounts for around 15% of the overall variance in the dataset and is highly impacted by two parameters: pH (0.74) and DO (0.71). DO is required for fish and other aquatic species to survive. In the appropriateness study, however, these characteristics were deemed to be within the safe limit.

With an eigenvalue of 1.33, it explains around 14% of the overall variance in the dataset. Nitrate (0.83), Pb (0.73), and BOD (0.73) are the most influential factors in this component (0.60). BOD, on the other hand, was found to be similarly loaded on PC1 (0.60) and PC4 (0.60). These exceeded the maximum permitted level during the rainy season and were thus deemed unfit for consumption. Especially for infants and pregnant women, nitrate levels in drinking water can be harmful to one's health. If the element lead in the water reaches the bloodstream, it generates a high blood lead level and hence excessive amounts of lead in UWBs might cause health problems. In addition to causing damage to the kidneys and brain, it can also prevent the production of red blood cells, which carry oxygen throughout the body.

Table 8 summarizes and presents the cluster diagnostic of the three easily recognizable groups. A group ID, sample ID, cluster ID, the proportion of samples in the cluster, group means, and standard deviation are all provided. Cluster I (Group A) accounts for 43% of the sample stations, according to Table 8, with a group mean of 25.2. Cluster II (Group B) is responsible for 14% of the spatial observation units, with a group mean of 21.5. Cluster III (Group C) accounts for 43% of the samples and has a low group mean of 16.04.

In terms of similarity, Group A and Group C are equally loaded in terms of similarity (representing 43 + 43 = 86% of the sample stations). The difference between the two groups, however, is striking, as seen by the mean values, even if variance within Group A is lower (S.D. 26.22859) than Group C (S.D. 34.42239), showing greater disparity within the group. Cluster II (Group B) is unusual in that it has only three examples and represents only 14% of the sample stations. Due to the limited number of objects, it has a medium mean value (21.5) and a lower SD (19.48524), indicating more homogeneity (less distance) within the group.

Table 9 displays the descriptive statistics (means and standard deviations) for the selected clusters (1–3). In comparison to Groups B and C, Group A appears to be the most affected by pollutant concentration. COD (91.6), Chloride (87.5), TSS (86.9), and BOD (37.2) are important contributing variables to Cluster 1 (Group A), as seen by their high mean values in the table. Cluster 2 (Group B) is substantially influenced by salinity (89.7), turbidity (65.0), EC (24.5), and TDS (18.3). Except for Chloride, which is significantly high, characteristics in Group C were shown to be modestly linked with Cluster 3 (73.5).

Cluster 1 (Group A): The mean value indicates that factors (COD 91.63), Chloride (87.55), and TSS (86.89) have a considerable impact on this group (Table 9). This collection of samples is found in CMC's western and eastern regions, mainly near the Bay of Bengal's coastal area and the Karnafulli River. Because tidal creeks and streams are connected, members of this group are affected by urban floods, tidal floods, and storm surges. The area suffers from high pollution concentrations because of the synchronization of surface runoff in the higher catchment areas, and river flows in the low-lying areas with that of periodic tidal inundation. There are other reasons as well; these locations are home to the Chittagong Export Processing Zone (EPZ), which houses medium and heavy businesses. As a result, contaminants make their way into UWBs and pollute them. Salinity (89.75), Turbidity (65.00), EC (24.48), and TDS all have a strong impact on Cluster 2 (Group B) (18.33). This group's members were discovered around CMC. Due to excessive surface runoff during the rainy season, the water bodies are impacted by urban floods. Furthermore, the urban slum dwellers near surrounding water bodies are negatively impacted by their unhealthy lives. Cluster 3 (Group C): This group discovered a significant link to Chloride (73.51). This group's members may be found in the east and south of CMC, although there are no apparent distribution trends. Since the samples were found at the Karnafully River's edge, they are thought to be linked to tidal creeks and streams. Some sampling sites are also strewn around the residential and business regions. Residential and commercial wastes are more likely to be the point sources of pollution in the water bodies included in this category.

Bangladesh is a tropical country with hot, humid summers and heavy rainfall during the rainy season, which causes floods in Chittagong. As a result of the increased activity of algae and microorganisms during the monsoon, surface runoff suffers from excess water, and turbidity rises. Other contaminants include sawdust, wood flames, plant residues, clay/mud, sand, silt, and chemicals in the water. Salinity rises when dissolved solids and salts accumulate in UWBs, increasing electrical conductivity; nitrate and turbidity concentrations have already exceeded the city's maximum allowable limit. Similarly, the maximum permissible concentrations of BOD and COD have been exceeded. BOD is produced in cities from a variety of sources, including leaves and woody debris, dead animals and plants, livestock manure, effluents from pulp and paper mills, wastewater treatment plants, feedlots, and food-processing enterprises, malfunctioning septic systems, and urban storm stormwater. COD-rich water includes high quantities of decomposing plant debris, human waste, or industrial effluent, all of which are harmful to aquatic life. Concisely, environmental deterioration in UWBs is manifested by the presence of dissolved solids (e.g. chloride); excessive concentrations of dissolved salts (e.g. nitrate), greater BOD and COD, and increased turbidity. The problem has a microbiological component as well; UWBs are heavily contaminated with TC and FC bacteria. The observed TC value during the rainy season was 1,100, compared to the normal value of 50 MPN-100 ml⁻¹, which is worrying. Disease-causing organisms (pathogens) may be present in the water system if they are found in the pathogens (disease-causing organisms) in drinking water. The inability to maintain a ‘closed’ system is the most common cause of TC bacteria in a water system. The majority of germs that might pollute surface water come from human or animal excrement. The primary sources of TC and FC bacteria in freshwater are failing septic systems, animal waste, and wastewater treatment plant discharges. However, extreme river flow, i.e. synchronization of tidal oscillations and heavy rains; inflow of water caused by the monsoon season, and flash floods are primarily responsible for the periodic inundation, particularly during the rainy season. Besides, flooding increases the rate of urban discharge via surface runoff, which permits dirty water (urban wastes and industrial fluids) to combine with standing water bodies. Furthermore, improper municipal waste management, unidentified blockage of municipal drains by urban solid wastes, illicit drain connection to water bodies, encroachment of tidal creeks, channels, and streams, low-lying topography of certain parts of the city, and saline intrusion from the Bay of Bengal are the main causes of urban floods in Chittagong. As a result, the water quality in UWBs quickly deteriorated.

The current research reveals that almost all water bodies in CMC are polluted with various contaminants and are unsafe for human consumption unless properly treated. Due to pollution levels, the drinking water quality of UWBs in CMC is found to be unfit for family consumption during the rainy season. Fresh water shortages also exist in the urban water supply system. However, the application of MATs is a suitable and advanced technique for the analysis of the quality of water, displaying sample patterns, and making clear assessments of the unabridged research. The SA analysis found that the mean value of the maximum sample has exceeded the allowable limit and they have a substantial relationship. The bulk of the characteristics employed in the SA were determined to be inappropriate for drinking during the rainy seasons. EC, BOD, COD, Turbidity, Nitrate, Lead, TC, and FC are the eight metrics that have exceeded the maximum allowed level. Besides, the PCA extracted four factors (PC1–4). These factors [Factor I (PC1), Factor II (PC2), Factor III (PC3), and Factors IV (PC4)] explain roughly 25, 19, 15, and 14% of the overall variation. With an eigenvalue of 10.23, the PCA indicates that the first four components (PC1–4) account for 73.1% of the total variation in the dataset. Moreover, the CA recognized three different groups or clusters (A, B, C – Clusters I–3) of the entire sample. In these clusters, 43% loaded in Group A, 14% loaded in Group B, and 43% loaded in Group C of the entire sample, respectively. These loading or mean values denoted greater disparity with Group A and less disparity with Group C. However, Group A (comprising nine members) is the most affected by pollution levels in CMC. The cause of this unacceptably bad situation is a combination of environmental, socioeconomic, and managerial issues. Because of the consumption of contaminated water, the health of city inhabitants, particularly the poorer parts of the population, is in danger; public and private sector awareness development on water contaminants is essential to enhance public healthcare delivery. Regular monitoring and assessment of a large number of water quality indicators are critical for efficient water quality maintenance through suitable control methods.

The author, M.H.M., acknowledges the Social Science Research Council, Planning Division, Ministry of Planning, the government of the People's Republic of Bangladesh, M.Phil. fellowship program for supporting his research.

M.H.M. – conceptualization, methodology, software, formal analysis, investigation, resources, data curation, writing – original draft, visualization. M.A.T.C., Md.M., and K.Md.B.A. – methodology, critical review, editing, formal analysis, and supervision. Md.H.R.B. – methodology, data curation, formal analysis, and supervision. S.D. A.J.M.M., and J.D. – methodology, data curation, and formal analysis. S.I. – methodology, formal analysis, resources, and editing. All authors read and approved the final manuscript.

This work received financial support from the Social Science Research Council, Planning Division, Ministry of Planning, the government of the People's Republic of Bangladesh (Ref.: 20.00.0000.309.02.815.12-116; Date: 19/01/2014).

Participation in this survey was voluntary, and respondents had the right to withdraw from the survey at any time during the interview. Before participating in the survey, potential respondents had to answer a yes/no question to confirm their consent to participate voluntarily. After receiving their verbal consent, respondents were requested to participate in the questionnaire survey.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.