Introduction

Adulteration of surface water quality as a consequence of natural or man-made activities has garnered global attention for their conservation and protection (Carpenter et al. 1998). The expulsion of untreated municipal and industrial waste, agricultural run-offs, leaching from landfills, mining and other activities have not only been hostile to aquatic ecosystems but also have introduced various trace elements such as toxic metals which have a persisting and non-biodegradable character (Carpenter et al. 1998; Sin et al. 2001). Extensive programs for water quality monitoring and assessment have thus surfaced worldwide so as to counter any activity causing the degradation of such resources.

Monitoring programs evaluate a broad array of physical, chemical and biological water quality parameters as well as the concentration of heavy metals in water. This necessitates the integration of these large and complex data sets into meaningful results that can represent the overall water quality status of a water body and can also be presented to planners and decision makers to take remedial action during an event of pollution. This led to the evolution of Water Quality Indices (WQIs) which aggregate a large set of measured parameters into a single numeric value (Zandbergen and Hall 1998). Nowadays, heavy metal contamination has turned out to be an area of major focus for water quality researchers due to its toxicity and abundance (Sin et al. 2001). Heavy metals can cause fatigue and damage the operations of brain, lungs, liver, kidney, blood composition and other important organs. Chronic exposure to heavy metals can damage the neural system paving the way for multiple sclerosis, Parkinson’s disease, and muscular dystrophy (Järup 2003; Harmanescu et al. 2011). Ingestion of lethal doses can cause cardiovascular collapse and renal tubular damage. These effects emphasize the need for quantification of heavy metals by quality indices such as the Heavy Metal Pollution Index (HPI) given by Mohan et al. (1996) and the Contamination Index (CI) developed by Rapant et al. (1995) and refined at Geological Survey of Finland. These assessments are necessary not only for evaluating the heavy metal contamination as a numerical score but also for assessing the scope of potability of water.

Apart from WQIs, the potential of multi-objective decision-making methods in stream restoration efforts has been evaluated by researchers in modifying WQI ranking, redressing management issues such as storage system, performance assessment, demand response and renewable energy sources (Aalami et al. 2010; Sianaki and Masoum 2013; Zahedi 2017; Yousefi et al. 2018). Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) is an effective methodology for ranking a number of conceivable alternatives by gauging their Euclidean distances (Sianaki et al. 2018). The TOPSIS method based on information entropy tries to arrive at a positive ideal solution (PIS) and a negative ideal solution (NIS), and then finds the scenario nearest to the PIS and farthest from the NIS (Sianaki et al. 2018).

WQIs are used for assessing the water quality with respect to physico-chemical parameters and heavy metal contamination is quantified by HPI and CI, as conventional WQIs do not include them in their computations. This necessitates the need for the TOPSIS method for not only providing an overall ranking of the sites by taking into account both the physico-chemical parameters and heavy metals but also prioritizing decisions in time of contingencies.

In this study, water quality of Manas River in Assam (India) has been assessed. Its water quality has been represented in terms of WQI. Heavy metal contamination was evaluated using HPI and CI. Health hazard due to heavy metals was expressed as hazard quotient (HQ), developed by U.S. Environmental Protection Agency (US EPA). Overall ranking in terms of pollution level of each sampling site was decided by TOPSIS and verified by cluster analysis (CA). Although, earlier studies have focused on modifying expected conflicts between drinking water quality index and irrigation water quality index, and validation of groundwater quality indices and its classes by the application of TOPSIS (Zahedi 2017; Zahedi et al. 2017; Yousefi et al. 2018), no study has identified relatively less polluted sites and relatively more polluted sites by removing conflicts between drinking water quality index and heavy metal pollution indices. Also, this study depicted the utility of the TOPSIS method in identification of relatively less polluted and relatively more polluted sites within the same cluster in cluster analysis.

Materials and methods

Study area

The Manas River has its origin in the Himalayan foothills between southern Bhutan and India. It is regarded as the largest river system of Bhutan and debouches into India through western Assam. It covers a distance of 104 kilometers in the state of Assam (India) before discharging into the Brahmaputra River at Jogihopa. The Manas River flows through the outskirts of Bongaigaon which is a major city in the state of Assam. As per Indian census of 2011, the population of the city is more than 1,00,000. The climate is humid sub-tropical in the region and it experiences the highest rainfall during the months of June and July (more than 350 mm). It receives less rainfall during winter (November to February). The average temperature of the city varies from 10 to 35 °C. The study area has been depicted in Fig. 1.

Fig. 1
figure 1

Map of water sampling sites in the Manas River

Full size image

Sample collection, preservation and analysis

The collection of water samples was done from nine sampling locations located along the stretch of the river (Fig. 1). All water samples were taken at 0.5 m below the surface of the river in triplicates. A total of 15 physico-chemical parameters were analyzed. pH and dissolved oxygen (DO) were measured in situ. Titrimetric method was done for the analysis of total hardness (TH) and total alkalinity (TA). Sodium (Na+), calcium (Ca2+), and potassium (K+) were analyzed using flame photometer. Anions were analyzed by ion chromatograph (IC). A total of six metals namely iron (Fe), manganese (Mn), chromium (Cr), copper (Cu), lead (Pb) and zinc (Zn) were measured in all the sampling sites by atomic absorption spectroscopy (AAS). Analytical procedures of Standard Methods for the Examinations of Water and Wastewaters 20th edition, published by APHA (2012) have been followed throughout the analysis.

Water Quality Index (WQI)

The method adopted for the calculation of WQI was in accordance with Alobaidy et al. (2010) and proceeds as follows:

Step 1 A total of 15 parameters were taken and each parameter was allotted a definite weightage (\({W_{\text{a}}}\)) according to its relative influence on the entire water quality varying from 1 to 5 (Table 1). Parameters which influenced more ominously the water quality were assigned a weight of 5 and the least influencing parameters were assigned a weight of 1 (Sharma et al. 2014). Relative weights (\({W_{\text{r}}}\)) were worked out using the formula given below:

Table 1 Relative weight of chemical parameters and their permissible limits
Full size table
$${W_{\text{r}}}={W_{{\text{a}}i}}~/\mathop \sum \limits_{{i=1}}^{n} {W_{{\text{a}}i}}$$
(1)

where \({W_{\text{r}}}\) and \({W_{{\text{a}}i}}\) denote the relative weightage and assigned weightage to each parameter, \(n\) indicates the number of parameters considered for the computation of WQI. The calculated value of \({W_{\text{r}}}\) for each parameter is given in the Table 1.

Step 2 A quality rating scale (\(Q\)) was calculated as:

$${Q_i}=\left[ {~{C_i}/{S_i}} \right] \times 100$$
(2)

In calculating the Q for the dissolved oxygen (DO) and pH, a different method was engaged such that the ideal values (\({V_i}\)) of pH (7.0) and DO (14.6) were subtracted from the measured values in the samples (Hameed et al. 2010).

$${Q_{i{\text{pH,DO}}}}=\left[ {~({C_i} - {V_i})/({S_i} - {V_i})} \right] \times 100$$
(3)

where \({Q_i}\) denotes the quality rating scale, \({C_i}\) denotes measured concentration of each parameter, and \({S_i}\) denotes the drinking water standard values for each parameter according to Bureau of Indian Standards (BIS 2012).

Step 3 Sub-indices (SI) were determined to calculate the overall WQI.

$${\text{S}}{{\text{I}}_i}={W_{\text{r}}} \times {Q_i}$$
(4)
$${\text{WQI}}=\mathop \sum \nolimits^{} {\text{S}}{{\text{I}}_i}$$
(5)

The computed WQI values were categorized in agreement with the suggested categorization of water quality (Yadav et al. 2010) as shown in Table 2.

Table 2 Water quality scale (Yadav et al. 2010)
Full size table

Heavy Metal Pollution Index (HPI)

The heavy metal pollution in both the rivers have been evaluated by two major indices (HPI and CI). HPI has been evaluated using the weighted arithmetic average method of indexing and was developed by Mohan et al. (1996) as follows:

Step 1 A unit weightage with a value inversely proportional to the recommended Si of the evaluated parameter was assigned.

Step 2 The HPI model proposed is given as:

$${\text{HPI}}=\frac{{\mathop \sum \nolimits_{{i=1}}^{n} {W_i}{Q_i}}}{{\mathop \sum \nolimits_{{i=1}}^{n} {W_i}}}$$
(6)

where \({W_i}\) represents the unit weightage of ith parameter; \({Q_i}\) denotes sub-index of the ith parameter and \(n\) indicates number of parameters included in the evaluation.

Step 3 Sub-index of the ith parameter was calculated as:

$${Q_i}=\mathop \sum \limits_{{i=1}}^{n} \frac{{\left| {{X_i} - {C_i}} \right|}}{{({S_i} - {C_i})}} \times 100$$
(7)

where \({X_i}\) and \({C_i}\) denote the monitored and ideal values of the ith heavy metal and \({S_i}\) denotes standard value of the ith heavy metal.

The numerator of the above equation (Eq. 7) indicates the numerical difference between the two values disregarding the algebraic sign. The critical value of HPI for drinking water is 100 (Prasad and Bose 2001).

Contamination Index (CI)

Individual components or parameters exceeding their upper permissible limits are aggregated for the calculation of CI for a sampling site. The combined effects of the toxic metals are thus summarized to a single numeric value. The scheme for calculation is as follows (Backman et al. 1998):

$$CI=\mathop \sum \limits_{{i=1}}^{n} {C_{fi}}$$
(8)

where,

$${C_{{\text{f}}i}}=\frac{{{C_{{\text{a}}i}}}}{{{C_{{\text{n}}i}}}} - 1$$
(9)

and \({C_{{\text{f}}i}}\) signifies the contamination factor of the ith parameter and \({C_{{\text{a}}i}}\) represents the analyzed value of the ith parameter. \({C_{{\text{n}}i}}\) is the upper permissible limit of the ith parameter.

Analytical values of components lower than their upper permissible limits were not taken into account. The grade scale of contamination index has been tabulated in Table 3.

Table 3 CI scale (Backman et al. 1998)
Full size table

Risk assessment on human health

The exposure, toxicity and risk assessment on human health included two major pathways namely ingestion and dermal absorption (US EPA 2004; Wu et al. 2009). The average daily dose (ADD) received from each individual pathway has been determined using equations modified from the US EPA.

$${\text{AD}}{{\text{D}}_{{\text{ingestion}}}}=\frac{{C \times {\text{IR}} \times {\text{EF}} \times {\text{ED}}}}{{{B_{\text{w}}} \times {\text{AT}}}}$$
(10)
$${\text{AD}}{{\text{D}}_{{\text{dermal}}}}=\frac{{C \times {\text{SA}} \times {K_{\text{p}}} \times {\text{ET}} \times {\text{EF}} \times {\text{ED}} \times {{10}^{ - 3}}}}{{{B_{\text{w}}} \times {\text{AT}}}}$$
(11)

where average daily dose ingestion (ADDingestion) and dermal absorption (ADDdermal) is calculated in µg/kg/day. The description of parameters involved in the calculation of ADDingestion and (ADDdermal) has been given in Table 4.

Table 4 Assumption parameters to derive the average intake values
Full size table

Risk characterization was enumerated by potential non-carcinogenic concerns calculated by hazard quotient (HQ). The estimation involved comparing the ADD of contaminants from each exposure pathway with the corresponding reference dose (RfD). The RfD values have been given in Table 5. Values of HQ exceeding 1 indicated concern of non-carcinogenic effects. The hazard index (HI) which is the summation of HQs from all possible pathways evaluates the total potential of non-carcinogenic risk from the water source.

Table 5 Reference dose of heavy metals
Full size table
$${\text{Hazard}}\;{\text{Quotient}}~\left( {{\text{HQ}}} \right)=\frac{{{\text{ADD}}}}{{{R_{\text{f}}}D}}$$
(12)
$${R_f}{D_{dermal}}={R_f}D \times AB{S_{GI}}$$
(13)

The values used in the equations have been obtained from the US EPA and the RfD values originate from the risk-based concentration table, US EPA (2004).

Cluster analysis

Cluster analysis (CA), a multivariate statistical technique groups objects on the basis of some particular characteristics they possess (Shrestha and Kazama 2007). Each and every object belonging to the same cluster possess some identical characteristics. Hierarchical clustering is the most common method of clustering and the resulting representation is done with the use of a dendrogram. In the present study, CA was executed on the data to evaluate the similarity among the sampling sites with respect to physico-chemical parameters as well as heavy metals. Hierarchical CA was implemented on the data by the means of Ward’s method and Euclidean distances were used as a measure of similarity (Güler et al. 2002; Yidana et al. 2008). The data sets were also standardized by ‘z scores’ so as to avoid any errors occurring from differences in data dimensionality and units of measurement. All the statistical analysises have been accomplished using the statistical package SPSS® (version 20.0 for Windows).

Technique for order preference by similarity to ideal solution (TOPSIS)

TOPSIS methodology based on information entropy aims at arriving at an alternative which is closest to the PIS and farthest from the NIS. It serves as an effective tool in decision-making processes and may be implemented as follows (Hwang and Yoon 1981):

Step 1 The “alternatives” (sampling locations) and the “criteria” (parameters) were specified for both the rivers to which the ranking was to be allocated according to their contamination status. Assuming the presence of “m” possible alternatives called \(A=~\{ {A_1}, \ldots ,{A_m}\} ~\) which are to be evaluated alongside “c” criteria \(C~=~\{ {C_1}, \ldots ,~{C_c}\}\).

Step 2 A matrix X was employed in assigning ratings to the criteria where xij indicated the value of alternative \({A_i}\) for criterion \({C_j}\)

$${X_{m \times c}}=\left[ {\begin{array}{*{20}{c}} {{x_{11}}}&{{x_{12}} \ldots .}&{{x_{1c}}} \\ \vdots &{~~~~~~ \ldots ~~~~.{x_{ij}}}& \vdots \\ {{x_{m1}}}& \cdots &{{x_{mc}}} \end{array}} \right]$$

Step 3 Criteria weights were calculated on the basis of information entropy techniques as follows:

$${q_{ij}}=\frac{{{x_{ij}}}}{{{x_{1j}}+ \cdots +{x_{mj}}}}~;~~\forall ~j \in \left\{ {1, \ldots ,c} \right\}$$
(14)

And,

$${E_j}= - \frac{1}{{\ln m}}\mathop \sum \limits_{{i=1}}^{m} {q_{ij}}\ln {q_{ij}};~\forall ~j \in \left\{ {1, \ldots ,c} \right\}~$$
(15)

where \(0 \leq {E_j} \leq 1\) where index with higher entropy has greater variation. Therefore, the weight of the criteria may be calculated as:

$${w_j}=~\frac{{{d_j}}}{{{d_1}+ \cdots +{d_c}}}$$
(16)

And, \({d_j}=1 - {E_j}\). All the weights were aggregated to a matrix \({w_{c \times c}}\).

Step 4 A normalized decision matrix was constructed (\({N_{m \times c}}\)) using vector normalization method as follows:

$${r_{ij}}=\frac{{{x_{ij}}}}{{\sqrt {x_{{ij}}^{2}+ \cdots x_{{mj}}^{2}} }}$$
(17)

Thus, \({N_{m \times c}}={\left[ {{r_{ij}}} \right]_{m \times c}}\).

Step 5 A weighted normalized decision matrix was constructed (\(V\)) as follows:

$$V={N_{m \times c}} \times {w_{c \times c}}$$

Step 6 The PIS and the NIS of the alternatives were computed as:

$${\text{PIS}}=\{ \hbox{max} {v_{ij}}|{v_{ij}} \in V\} =(v_{1}^{+}, \ldots ,v_{c}^{+})$$
(18)
$${\text{NIS}}=\{ \hbox{min} {v_{ij}}|{v_{ij}} \in V\} =\left( {v_{1}^{ - }, \ldots ,v_{c}^{ - }} \right)~$$
(19)

Step 7 The Euclidean distance of each alternative from the PIS \((d_{i}^{+})\) and NIS (\(d_{i}^{ - }\)) were calculated as:

$$d_{i}^{+}=\sqrt {\mathop \sum \limits_{{j=1}}^{c} {{\left( {{v_{ij}} - v_{j}^{+}} \right)}^{2~}}}$$
(20)
$$d_{i}^{ - }=\sqrt {\mathop \sum \limits_{{j=1}}^{c} {{\left( {{v_{ij}} - v_{j}^{ - }} \right)}^2}}$$
(21)

Step 8 Proximity or closeness coefficients (CC) of each and every alternative was calculated as:

$${\text{CC}}_{i}^{+}=\frac{{d_{i}^{ - }}}{{d_{i}^{ - }+d_{i}^{+}}}$$
(22)

Step 9 The alternatives were finally ranked according to their closeness coefficients.

Results and discussions

Descriptive statistics of monitored parameters

The descriptive statistics of the monitored physico-chemical parameters at a total of nine sampling locations of the Manas River have been shown in Table 6. From Table 6, it was observed that the pH of the river was within the guidelines provided by BIS of 6.5–8.5. A significant amount of chemical reactions occurring in nature are pH-sensitive, and there is a high influence of pH on the biotic compositions of aquatic systems. DO concentrations of the river were considerably high which may be ascribed to temperature variations and phytoplankton growth. TDS of the river was within the permissible limits of BIS guidelines (500 mg/L). The presence of high concentration of ions capable of carrying electrical charge contribute mainly to high EC values in rivers. TH and TA of the river were in their desirable limits. According to BIS guidelines, the desirable limits of both the parameters are 200 mg/L. The BOD5 values at majority of the locations were found to be within the desirable limits except in SSMR 5 (12.75 mg/L). The major cations and anions analyzed have been depicted in Table 6 for the river.

Table 6 Descriptive statistics of physico-chemical parameters of water samples collected from Manas River
Full size table

Water Quality Index

The WQI adopted for the evaluation of water quality of river was in accordance with Hameed et al. (2010). The results of the evaluation showed the fairly different grades of water quality along the stretch of both the rivers. The gradation of water quality of river has been done as per the quality scale given by Yadav et al. (2010). The evaluation of the WQI yields significant results which have been shown in Fig. 2. The calculated WQIs of the Manas River was found to be in the range of 54.3–91.1. The sampling locations SSMR 5 and SSMR 6 of the Manas River were graded as “very poor” by the quality scale of Yadav et al. (2010). The remaining sampling locations along the river were of “poor quality”.

Fig. 2
figure 2

Variation of WQI at all sampling locations

Full size image

Quantification of heavy metal contamination and risk assessment on human health

The concentration of heavy metals at the sampling locations for the Manas River have been depicted in Table 7. The results obtained from evaluating both the indices (HPI and CI) have been depicted in Table 8. The HPI of the sampling locations in the Manas River were in the range of 23.93–55.39. The water quality at all the nine sampling locations was graded to be suitable for human consumption, as the HPI values at all the locations were below the critical limit (≥ 100). The highest HPI value (55.39) was noted at the sixth sampling location of the Manas River which is located near Bongaigaon (Assam). At this location, the concentrations of the heavy metals were in the order of Fe > Cu > Zn > Mn > Cr > Pb. The results on evaluation of the contamination index of each location depicted that the fifth and the sixth sampling location of the Manas River having CI values of 2.32 and 2.07 were “moderately contaminated” or “medium polluted”.

Table 7 Descriptive statistics of heavy metals of water samples collected from Manas River
Full size table
Table 8 HPI and CI of Manas River
Full size table

From Table 8, it was observed that the indices (HPI and CI) contradicted each other. The HPI indicated that the water quality at all the locations of the Manas River was suitable for human consumption and the CI graded the fifth and the sixth sampling location of the river as moderately contaminated. This initiated the need for assessing the risk of heavy metal contamination on human health so as to provide a transparent picture of the potential risk of heavy metal intake to communities depending on the water of the Manas River for their day to day activities. The ADDingestion, ADDdermal, HQ and HI presented in Table 9 evaluates a comprehensive risk assessment to the human population residing along the stretches of the Manas River. The HQ ingestion and HQ dermal of all the heavy metals were below unity suggesting that the metals posed little or no health hazards when they enter through both the pathways. Furthermore, the overall hazard index (HI) of all the heavy metals came out to be well below unity suggesting the same. Although the risk characterization and assessment in this study has been done with utmost precision, there exist several uncertainties associated with the risk assessment which was emphasized by US EPA and other references. However, the results form a foundation on which several thorough investigations can be built upon for better assessments.

Table 9 Hazard quotient and hazard index of Manas River
Full size table

Cluster analysis and TOPSIS

In the study, WQI included the physico-chemical parameters and the HPI and CI included the heavy metal concentrations. However, characterization of the sampling locations became difficult as the results of the three indices contradicted each other. A weak correlation was found among them. Hierarchical cluster analysis (HCA) implemented on the data included both the physico-chemical parameters and heavy metals to group the sampling locations possessing similar characteristics. The visual representation of the clusters in the form of a dendrogram has been depicted in the Fig. 3. The consequence of the cluster analysis resulted in two major clusters. The first cluster represented the relatively less polluted sites (LP) and the second cluster grouped the more polluted sites (MP). The second cluster included the fifth and the sixth sampling locations (SSMR 5 and SSMR 6, respectively) which had the highest WQI as well as the highest CI. The overall ranking of the sampling locations given by the TOPSIS methodology has been shown in the Table 10. The TOPSIS method also ranked these locations as the most polluted sites by giving them an overall rank of 8 and 9, respectively. The first cluster had two sub-clusters containing the remaining sampling locations. The identification of the relatively more polluted sub-cluster or the relatively less polluted sub-cluster becomes difficult in such circumstances where WQI, HPI and CI are in dispute among themselves. However, in such a complicated scenario, the TOPSIS method clearly identifies the relative pollution level by giving them their overall ranks. The first sub-cluster including the sites SSMR 7, SSMR 9, SSMR 3 and SSMR 8 were ranked 4, 5, 6 and 7, respectively. The second sub-cluster including sites SSMR 4, SSMR 2 and SSMR 1 were ranked 1, 2 and 3, respectively. It can be inferred that the first sub-cluster is relatively more polluted than the second sub-cluster. The TOPSIS method provided an overall ranking to the sampling sites of both the clusters as well as proved efficient in ranking the sites within the sub-clusters.

Fig. 3
figure 3

Dendrogram of CA of Manas River

Full size image
Table 10 Closeness coefficients (CC) and TOPSIS ranks of all the locations of Manas River
Full size table

Conclusions

In this study, the surface water quality assessment of Manas River, Assam has been done using three quality indices, Hameed’s WQI, HPI and CI. The WQI graded two sampling locations near Bongaigaon as “very poor” and all other locations as “poor”. HPI of all the locations were below the critical value of 100, but the CI depicted that the two locations near Bongaigaon are “moderately contaminated”. Cluster analysis grouped the sampling locations in two major clusters LP and HP. TOPSIS was performed including all the measured parameters for characterization of sampling locations and provided an overall ranking of the sampling locations on the basis of their relative pollution levels. Furthermore, TOPSIS also served efficient in prioritizing sampling locations within the same cluster which was not possible to discern from the numerical scores of WQI based on physico-chemical parameters, and HPI and CI which included only the heavy metals. It was concluded that TOPSIS served as an effective tool in prioritizing decisions for policy makers based on these overall ranks for better water resources management and effective implementation of stream restoration strategies.