Multivariate statistical analysis to assess the surface water quality of a snow and glacier-fed river: A case from Alaknanda River basin

Anthropogenic activities have substantially impacted river systems, leading to a decline in water quality, a reduction in water availability, and a reduction in the capacity of aquatic life (Whitehead et al. 2009; Sharma et al. 2019; Sofi et al. 2022). Rivers contribute significantly to the sustainable socio-economic growth of a nation and makeup 0.006% of the global freshwater resources (Shiklomanov 2000; Grimmet al. 2008). With the exponential increase in population, tourism, agriculture, and urbanization, the Himalayan rivers are under extreme stress (Singh et al. 2020). The quality and quantity of rivers are seriously threatened by these human consequences and the ambiguous implications of climate change (Sharma et al. 2021). However, the water quality in high-altitude fluvial systems is significantly decreased by natural processes, including intense precipitation, rock or snow avalanche, glacier lake outburst, soil erosion, and agricultural activities (Kumar et al. 2020). Most of the population in the Himalayan area live along riverbanks and rely heavily on groundwater and river water for drinking, agriculture, and other household needs (Rani et al. 2021; Sofi et al. 2023). Poor waste management methods are in these places, particularly along the banks, and untreated or partially treated sewage is discharged into rivers, which subsequently impacts groundwater quality nearby (Sarkar & Paul 2016).

The change in the climate largely influences the movement of surface runoff via variability in the precipitation intensity and profound glacier retreat (Rani & Sreekesh 2019; Dubey & Goyal 2020). Furthermore, fluctuations in surface runoff, interflow, groundwater flow, and pumping in and out contribute significantly to river flow and the concentration of pollutants (Mirzaei Aminiyan et al. 2016; Thakur et al. 2018). A significant loss in snow and glacier retreat in the Himalayan regions has resulted in a freshwater shortage (Rani & Sreekesh 2019; Rautela et al. 2022a). Many non-natural variables affect river water quality in the Himalayan region, such as unprecedented urbanization, increasing tourism, shifting land use patterns, trash dumping, and agricultural runoff (Thakur et al. 2018). Water scarcity occurs downstream of the Himalayan rivers due to changing climatic conditions and other anthropogenic activities (Sati 2020). Therefore, identifying probable causes of impaired surface water quality is the foundation and requirement for water quality management (Tomar et al. 2022). In the headwater rivers, the installation of micro or small hydroelectric power plants with pondage can modify the flow characteristics, influencing its hydrodynamics and potentially altering sediment transport patterns, but it does not fundamentally transform a river into a lacustrine ecosystem (Sofi et al. 2022). Assessing water quality is imperative as it provides essential insights into the ecological health of these ecosystems and ensures the safety of water resources for both human and environmental needs. Multivariate statistical analysis plays a crucial role in this assessment by enabling the comprehensive evaluation of various parameters, helping identify patterns, correlations, and potential sources of pollution, which are instrumental in formulating effective strategies for maintaining and improving the water quality of these vital river systems (Kwan & Jo 2023).

In the field of traditional statistics, multivariate statistical analysis has become a well-established and powerful method for studying data from multiple variables at once. It is possible to evaluate the statistical laws governing relationships between many objects and various indicators using techniques such as cluster analysis (CA) (Corporal-Lodangco et al. 2014), discriminant analysis (DA) (Oda et al. 2020), principal component analysis (PCA) (Yang et al. 2013), and factor analysis (FA) (Aristizabal et al. 2019), when they are associated with one another. When it comes to chemical and physical measurements with multiple components, multivariate statistical analysis effectively reduces and interprets the data (Nnorom et al. 2019). It serves as a useful tool for identifying elements and sources that could influence water quality and aquatic systems (Chattopadhyay et al. 2012).

Monitoring surface and groundwater concerning WQI has been the subject of numerous studies (Bouza-Deaño et al. 2008; Wu et al. 2017; Nnorom et al. 2019; Valentini et al. 2021). It is not easy to understand and explain meaningful data from water chemistry due to several obscure human and natural factors (Pehlivan & Arslan 2007; Rani et al. 2021). In the above context, the present study uses a compound data set to generate a water quality index (WQI) to evaluate and assess the differences in overall quality between study sites and seasons. To compare drinking and domestic water standards with those provided by the Bureau of Indian Standards (BIS) and the World Health Organization (WHO), water samples were collected in four seasons (summer, monsoon, winter, and spring) from different sites from different altitudes. In addition to WQI and hydro-geochemical analyses, multivariate statistical techniques were used to understand complicated water quality grids to find solutions to pollution issues. In this way, the results of the study will provide important information about the river Akaknanda's surface water quality and contribute to thoughtful and sustainable water quality management.

The Uttarakhand state is divided into three main stratigraphic zones as outer or sub-Himalayan zone, the central or lower Himalayan zone, and the higher Himalayan zone, which is formed by largely tertiary sediments, granite, and crystalline rocks of un-fossiliferous sediments and highly fossiliferous sediments, respectively (Panwar et al. 2017). Regarding geology, the Alaknanda basin is separated into three lithological sequences: the Tethyan Sedimentary Sequence, higher Himalayan crystallines, and lesser Himalayas in the upstream, middle, and downstream regions, respectively (Panwar et al. 2017). The primary rock formations of the Alaknanda River are granite, schist, dolomite, quartzite, phyllite, gneiss, limestone, and migmatites (Bickle et al. 2003). The primary structural faults across the Alaknanda basin are the Alaknanda fault, Main Central Thrust (MCT), North Almora Thrust (NAT), and South Tibetan Detachment System (STDS). The Alaknanda River basin consists dendritic drainage pattern with very high drainage density in the upstream sections of the river and its tributary. The river travels a total distance of 195 km with steep terrain and significant flow. Hydrologically, the Alaknanda River shows much diversity in terms of headwater contribution. The tributaries of western Dhauliganga, Nandakini, Pinder, and Mandakini contribute a significant amount of snow and glacier melt to the Alaknanda River. Previous studies show that snow and glacier melt contributes 23–38% of total river flow. The observed streamflow at the Devprayag station ranges from 50 to 4,400 m³/s (Rautela et al. 2022a). Also, at Devprayag gauging station, the Alaknanda River carries three times more total suspended sediments (TSS) than the Bhagirathi River (Panwar et al. 2017).

The water samples were collected from the 15 sites (mainly at Prayags) situated along the Alaknandariver and its tributaries (Figure 1). Three samples were taken from the river at each site during four seasons (monsoon, winter, spring, and summer) for the years 2021–2022. Samples were taken in 500 mL sterilized polyethylene bottles that had been cleansed with metal-free soap, rinsed with distilled water, and then immersed in 10% nitric acid for 24 h. To prevent changes in water quality and interference, all of the samples were maintained at 4°C. Water samples were collected, preserved, transported, and analyzed using the techniques outlined in APHA-AWWA-WEF (1998). To keep the samples free from contact with the atmosphere, the polyethylene bottles used to store the samples were tape-sealed.

After samples were preserved, the physicochemical parameters were assessed in the laboratory according to the Bureau of Indian Standards (BIS 1991) and the American Public Health Association (APHA) (Eaton et al. 2005). For measuring pH, turbidity, dissolved oxygen (DO), and electrical conductivity (EC), we used pH meters, nephelometers, and conductivity meters (PC-II, Hach, USA). The laboratory also assesses other parameters using volumetric analysis.

Vi represents estimated values for the parameter, and V0 represents the parameter in ideal water conditions, generally considered zero except for pH and DO, which are 7 and 14.6 mg/L, respectively. The rating scale used in this study is presented in (Table 1).

Freshwater, marine water, and sediment quality have all been characterized using multivariate statistical techniques (Noori et al. 2010; Hamid et al. 2016). A multivariate statistical approach has recently been widely used to assess water quality (Kim et al. 2005; Barzegar et al. 2020). The correlation coefficient between two variables is calculated using the correlation analysis approach. The strength and importance of the two variables can be used to gauge the link between them. Correlation (r) indicates the strength, while probability levels show the importance (p values). Stronger relationships have higher correlation coefficients, but more significant relationships have lower p-levels.

PCA is a statistical method that uses a group of variables to extract linear relationships between them (Simeonov et al. 2003). PCA allows data analysis without sacrificing much information (Singh et al. 2005). According to the principal components (PCs) produced during the analysis, principal component 1 (PC1) explained the most variance in the original data. The factor loadings were classified as ‘strong,’ ‘moderate,’ or ‘weak’ based on absolute loading values of 0.75, 0.75–0.50, and 0.50–0.30 (Vieria et al. 2012), but loading does not indicate a component's relevance; rather, it indicates the relative significance of a variable within the component.

The water quality indices were calculated from the physicochemical parameters obtained from the 15 sampling locations over four seasons (monsoon, winter, spring, and winter) with their average values and permissible and desirable limits (Tables 2 and 3). The average values of water quality parameters are below the desirable limits except for turbidity and TSS. The average values of TSS range from 0.04 to 9.36 g/L. The highest TSS has been measured in the Alaknanda River at the Karnapyarag site. However, the Pinder river contributes a significant amount of suspended solids to the Alaknanda River. There is an effect of storage reservoir also seen in the amount of TSS. The Mandakini River at Rudraprayag and Alaknanda River at Srinagar show less TSS than the other sites due to Singoli-Bhatwari and Alaknanda Hydro Electric Power plants. The higher concentration of suspended solids in the river water may affect the mechanical components of the turbine and cause wear and tear (Rautela et al. 2022b).

The pH values of all the sampling sites and seasons range from 7.1 to 8.25. The pH value is well correlated with the TSS. Similar sampling sites, as discussed above, have higher pH values. Higher pH levels influence aquatic life, mucous membranes, corrosion, and the flavor of water (Narasimha Rao et al. 2011). The DO in the sampling sites ranges from 5.25 to 10.26 mg/L. The sites at higher altitudes (Mana and Kedarnath) have significantly less DO than the downstream sites. However, the downstream regions have to DO more than 6 mg/L, which is desirable for aquatic animals. The significant amount of DO is the adequate dissolution of oxygen during river water's up and down movement. Throughout the investigation, EC accounted for 100% of the desired range, with mean values ranging from 92 to 156 μS/cm. The EC is somehow increased during the summer and monsoon seasons due to the melting of glaciers. The trends for EC measurements of river water samples at Pinder (Karnapyarag) and Alaknanda (Karnaprayag) are found to move parallel to each other, with higher values throughout the study interval.

Hardness is a characteristic of water that hinders soap from lathering properly. The main cations that contribute to hardness are calcium and magnesium. Strontium, iron, and manganese are additional cations that contribute to hardness. Carbonate and bicarbonate anions make up the majority of the anions that cause hardness. The mean concentration of hardness in the sampling sites ranges from 25 to 89 mg/L. The TH of the all-sampling sites falls within the desirable limits. Although water hardness alone has no harmful effects on health, a higher quantity of hardness in the water can lead to heart disease and kidney stone issues (Napacho & Manyele 2010). The Total Alkalinity (TA) of all the sampling sites ranged from 28 to 97 mg/L and fell within the desirable limits. Alkalinity in water is caused by carbonates, bicarbonates, and hydroxides, which might be dissolved sediments, salts, or rocks (Krishna Kumar et al. 2012). The presence of suspended solids, including clay, silt, colloidal organic matter, plankton, and other creatures, causes turbidity, which is a measurement of the purity of the water (Dorner et al. 2007). The turbidity of the sampling sites ranged from 3 to 18.1 NTU. During monsoons and summer, the samples' turbidity exceeds the desirable limit. However, the Pinder River shows very high turbidity during both summer and monsoon season and deliver a significant amount of suspended sediments to the Alaknanda River. The turbidity and TSS have been largely influenced by the melting of debris and moraine-covered glaciers and intense rainfall in the summer and monsoon seasons. TSS and turbidity are at the desirable limit during the winter and spring season, where the melting of snow and glaciers are negligible. The overall water qualities showed significant variation during monsoon and summer, maybe due to increased sediment transport during the season. The suspended solids content is the only factor seen beyond the permissible limit for most cases, impacting the water quality in the sampling sites.

The substantial link among the measures was developed by statistical analysis using Pearson's correlation coefficient between several river water quality metrics (Bhandari & Nayal 2008; Joshi et al. 2009). The data analysis produced an R-value, which is a correlation showing the linear relationship between the data pairs. A linear relationship indicates that if one variable rises or falls, the other one does so linearly as well. Correlation coefficients with values approaching +1 (positive correlation) indicate that while one variable rises, the other rises almost linearly. On the other hand, a correlation value that is close to −1 indicates that when one variable rises, the other falls almost linearly. Values approaching 0 denote a negligible or nonexistent linear connection between the variables (Mudgal et al. 2009). When data are unrelated, there is no connection between the data points. In Table 1, there is a strong positive correlation shown between the TSS-turbidity (r = 0.915), TA-TH (r = 0.896), TSS-TA (r = 0.726), and TSS-TH (r = 0.827) during the monsoon, winter, spring, and summer season, respectively (Table 4).

Moreover, there is also a less or negative correlation found between TA and DO (r = 0.402), TSS and TA (r = −0.367), EC and DO (r = −0.507), and TA and DO (r = 0.451) during monsoon, winter, and spring and summer season, respectively (Table 4). TSS and turbidity are strongly correlated with each other during the summer, spring, and monsoon seasons, whereas in the winter, there is no correlation between these two parameters. The results also indicate that the river has a significant amount of TSS concentration in the high flow period of the Alaknanda River and its tributaries, and the melting of snow and glaciers will affect the amount of TSS and turbidity of the riverine flow dynamics. The strong and weak negative correlation between the measures might be interpreted as the main causes of seasonal fluctuations in water quality.

In the monsoon season, PC1 explains approximately 3/4th, i.e., 75.30% of the total variance, and has a moderately positive correlation with pH; Turbidity and TSS suggest the natural weathering sources that affect the water quality parameters (Table 6). At the same time, PC2 and PC3 are insignificant in the monsoon season. In the winter season, PC1 explains 40.70% of the total variance, and a moderately positive correlation with DO, TH, and TA suggests the basin's lithogenic factors influence the water quality (Table 6). However, during the winter, the PC2 and PC3 hold significant variance values. PC2 explains 33.8% of the total variance. Moderately positive and negative correlated with the EC and turbidity.

In comparison, PC3 explains 15.9% of the total variance and is negatively correlated with the turbidity and TSS, respectively, showing that in the winter season, the water flowing in the river and its tributaries will not be much affected by the suspended sediments (Table 6). In the spring season, PC1 explains 46.8% of the total variance and is moderately positively correlated with the DO, turbidity, and TSS, respectively, indicating the significant contribution of the TSS in the riverine flow. On the other hand, PC2 explains 18.90% of the total variance and moderately positively and negatively correlated with the EC and TA, respectively, while PC3 explains 18.10% of the total variance and moderately correlated with the EC and TA (Table 6). In the summer season, PC1 explains 70.30% of the total variance and is moderately positively correlated with the turbidity and TSS, respectively, while PC2 and PC3 are insignificant in the summer seasons because of the eigenvalue less than 1 (Table 6). In the summer, the turbidity and TSS affect the river's water quality, as we discussed earlier, the possible causes of the availability of TSS.

The current study clearly illustrates the use of multivariate statistical approaches in conjunction with WQI with graphical representation. Such instruments are used in the current study to comprehend the water quality parameters of the surface water of Alaknanda and its tributaries. The parameter values of all the water samples analyzed were well within the ideal ranges defined by BIS (1991) and WHO (2011), except somehow, TSS and turbidity of the water samples increased in the summer and monsoon season due to the intense glacier melting and erratic rainfall pattern. Except for all spots in summer and monsoon season where the TSS and turbidity were outside the acceptable range, the overall water quality is thus safe for domestic use with proper sedimentation. The higher suspended sediment concentration in the river water will negatively impact the aquatic biodiversity and runoff of river hydroelectric power plants in the downstream regions. According to the WQI, most water samples fell into the bad category throughout the summer and monsoon seasons, followed by the good category during the summer and spring. The PCA and correlation matrix highlight the key anthropogenic variables, such as agri-runoff and household wastewater runoff that are the primary influences on the water quality indicators. These factors include rock–water interaction, ion exchange, and the leaching of parent materials. The current study's findings will provide baseline information to the municipal corporations and planners of the cities situated at the river banks.

The authors thank the Director, Govind Ballabh Pant Institute of Engineering and Technology, Pauri Garhwal, Uttarakhand, and the Head of the Civil Engineering Department, Indian Institute of Technology (IIT) Indore, for providing facilities to make the present study possible. The research work was conducted as a part of the research project entitled ‘Assessment of Flood Vulnerability in upstream catchments of Himalayan River Basin,’ sponsored and funded by the Indian Space Research Organization (ISRO), SAtellite based RIver hydrological Techniques and Application (SARITA) program, Govt. of India, New Delhi is thankfully acknowledged. The authors thank Mr Sugato Panda (Research Scholar, IIT Indore) for assisting in the Analysis of WQI. The authors also thank Mr Lal Singh for their continuous efforts, measurement of field data, and assistance during the fieldwork.

All authors have read, understood, and have complied as applicable with the statement on ‘Ethical responsibilities of Authors’ as found in the instructions for authors and are aware that with minor exceptions, no changes can be made to authorship once the paper is submitted.

K. S. R. and D. K. conceptualized the whole article. K. S. R. and D. K. developed the methodology, K. S. R. and D. K. rendered support in formal analysis and investigation. K. S. R. wrote the original draft, K. S. R., D. K., B. G. R. G., A. R., B. S. K. and A. K. D. wrote the review and editing; D. K. and A. K. D. supervised the article.

Indian Space Research Organization (ISRO), SAtellite RIver hydrological Techniques and Application (SARITA) program provided financial support for this study. The opinions expressed herein are those of the authors and do not necessarily reflect the views of the study sponsors.

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

The authors declare there is no conflict.