Hydrochemistry and dissolved solute load of meltwater in a catchment of a cold-arid trans-Himalayan region of Ladakh over an entire melting period

Major Himalayan rivers such as the Ganga, Brahmaputra and Indus carry a significant contribution from snow and glacial melt (Bahadur et al. 1978; Pandey et al. 1999; Kaser et al. 2010; Wulf et al. 2010). A recent study using an isotopic technique shows that the glacial melt contributes to 15% of the total runoff in the Gaerqu River catchment, located in the source region of the Yangtze River, China (Liu & Yao 2016). Another related study shows a poor relationship of glacial discharge with suspended sediment concentration due to hysteresis effect in glacial melt stream in Gangotri Glacier (Arora et al. 2014). A variation in glacial and periglacial area influencing the water quality was observed in glacier originating rivers (Jones & Parker 2015). These studies indicate that melting, discharge, land use change and related factors influence the water quality changes in glacier regions.

Meltwater samples were collected from South Pullu station of Ganglass catchment during the period of May–September 2010. The samples were collected once daily at high flow period (1,700 h) in a 300 ml prewashed polyethylene bottle. After being taken to the laboratory in cold conditions, electrical conductivity (EC) and pH of the water samples were measured using a portable multi-parameter meter (HACH-Sension 156). Suspended matter from the sample was removed through a filtration process using 0.45 μm Millipore membrane filters. Bicarbonate was measured by acid titration method with a semi-auto titrator (877 Titrinoplus) using 0.01M HCl as titrant and 4.6 pH as the end point. Major cations (Ca, Mg, Na and K) were analysed using atomic absorption spectroscopy (Thermal Scientific M series). The other analysis used different standard methods such as mercury (II) thiocyanate method (Florence & Farrar 1971) for chloride, brucine-sulphanilic acid method for nitrate, molybdosilicate method (APHA 2005) for dissolved silica, turbidimetric method for sulphate and ascorbic acid method (APHA 2005) for phosphate, respectively. Discharge was measured at South Pullu using the area–velocity method. Charge balance error, calculated using the formula {(TZ⁺ − TZ⁻)/(TZ⁺ + TZ⁻)} × 100 is <10%, confirming reliability of the data.

The snow melt period is segregated into early snow melt (May–June), peak snow melt (July–August) and late snow melt (September) periods to evaluate the seasonal variations in the meltwater chemistry. To identify the sources and processes controlling meltwater chemistry, the data have been plotted against each other in different ways. C-ratio is the ratio of HCO₃ to (HCO₃ + SO₄), which is used to identify the major proton producing reactions, i.e., carbonation and sulphide oxidation (Brown et al. 1996). A ratio of 1.0 indicates weathering by carbonation reactions (Reynolds & Johnson 1972) and a ratio of 0.5 represents coupled reaction of both with proton derived from pyrite oxidation. To identify the main water types, major ion compositions were plotted on piper plot (Piper 1944) using AquaChem software. Further, the use of statistical tools like Carl Pearson correlation (significant level 0.01 and 0.05) and factor analysis have provided better understanding of the solute acquisition process in the system. R mode factor analysis has been used considering factors with eigen values >1 and indicating association between variables and factor by its loading. TDS flux has been calculated using the discharge and TDS data of meltwater, while cation weathering rate for the catchment has been calculated using dissolved cations concentration (Ca, Mg, Na and K), discharge data and catchment area (Singh & Hasnain 1998; Singh et al. 2014).

Runoff in the upper Ganglass catchment is generated mainly from the snow melt. A summary of data during these periods is provided in Table 1. The slightly acidic to neutral pH values (5.2–7.3) of the runoff resembles the snow meltwater from the Kashmir valley (Lone & Khan 2007). EC of the meltwater is highest during the late snow melt period (50.99 ± 23.18 μs/cm), followed by early (33.93 ± 9.30 μs/cm) and peak snow melt period (26.15 ± 6.19 μs/cm). Ca and HCO₃ are the most dominant ions present throughout the melt period. Cation shows a similar trend of Ca > >Mg > Na > K for the whole melt period. The anions show a regular trend of HCO₃ > >SO₄ > Cl > NO₃ > PO₄ for peak and late snow melt periods. A similar trend has been observed in surface water of SE Kashmir (Jeelani et al. 2011; Bhat et al. 2014). Deposition of alkaline dust causing alkalinity and high Ca values in the fresh snow samples of Kashmir valley has also been reported (Lone & Khan 2007).

Correlation matrix (Table 2) and factor analysis (Table 3) add further insights to these results as they identify other processes in addition to carbonate weathering which are regulating the meltwater chemistry of the upper Ganglass catchment. The strong correlation between Ca, Mg, K, HCO₃ and EC with each other during the early melt period indicates the role of ion exchange process as well. Most of the catchment experiences seasonal freezing in winter with very low winter temperatures up to −27.4 °C. Hence, meltwater infiltrations to the surface soil also help in seasonal thawing and enhance the residence time of snow meltwater in the catchment, which help it to acquire more ions. The significant negative correlation of SO₄ with EC, cations and HCO₃ suggests inputs from multiple sources (Khadka & Ramanathan 2013). Atmospheric deposition of SO₄ from burning of coal and oil for winter-time heating might be the other possible source. During the peak melt period, all the major cations were showing significant correlation with each other suggesting input from combined carbonate/silicate weathering and ion exchange process (Oinam et al. 2012). Significant correlation of silica with all major cations and HCO₃ indicates weathering of aluminosilicate minerals (Singh et al. 2011). The correlation matrix shows the dominance of carbonate weathering during the late melt period. In the factor analysis, the early melt period explains 76% of the total variance with four major factors showing an eigen value >1. Factor 1 with very high loading (>0.8) of K, Ca, Mg, HCO₃ and EC explains about 43% of variance indicating carbonate weathering followed by ion exchange processes. Factor 2 (12% of variance) shows high loading of silica and PO₄ indicating silicate weathering along with phosphorite leaching or anthropogenic contribution (Singh et al. 2014). Factor 3 (10.6% variance) with high loading of Na and Cl suggests deposition of sea salt aerosol. For the peak melt period, a total of three factors explaining 68% of the total variance have been identified. Factor 1 (47.5% variance) has high loading of Na, K, Ca, Mg, silica, Cl, HCO₃ and EC explaining carbonate and silicate weathering along with some atmospheric deposition. Factor 2 (11.3% variance) with high loading of NO₃ and pH shows atmospheric deposition (Singh et al. 2014), while factor 3 (9.4%) with high loading of PO₄ and SO₄ shows an anthropogenic contribution (Anshumali & Ramanathan 2007). The late melt period identifies four factors explaining 66% of the total variance. Factor 1 (33% of variance) shows high loading of Na, Ca, Mg, HCO₃ and EC indicating the contribution from carbonate weathering and ion exchange process. Factor 2 (13% of variance) with high loading of Na, silica and SO₄ suggests sulphide oxidation coupled with silicate weathering. Factor 3 (10.9% of variance) with a high loading of K and silica indicates silicate weathering. Factor 4 (9% of variance) with a strong loading of PO₄ and Cl shows the anthropogenic and atmospheric contribution. Thus, the overall factors for the melt period can be grouped as carbonate and silicate weathering, ion exchange process and atmospheric deposition (alkaline dust, anthropogenic pollutants and sea salt aerosol). The atmospheric deposition can be linked to significant snow fall (which persisted until June 22, 2010) observed in the basin causing a higher accumulation of snow.

The average TDS flux calculated for the early, peak and late melt period for the upper Ganglass catchment is 0.64 t day⁻¹, 3.02 t day⁻¹ and 1.31 t day⁻¹, respectively. This is much lower when compared with TDS flux from glaciers in the monsoon regime. For example, Gangotri glacier has TDS flux of 310.9 t day⁻¹, 325.3 t day⁻¹ and 113.4 t day⁻¹ for the pre-monsoon, monsoon and post-monsoon season, respectively, and the value ranges from 2.77 to 31.32 t day⁻¹ for Dokirani glacier (Hasnain & Thayyen 1999; Singh et al. 2014). This significant difference between the monsoon and cold-arid regime (present study) is due to the smaller glacial area as well as very low precipitation in the latter system, resulting in low discharge (0.4–2.15 m³s⁻¹), and hence, lower TDS flux. Compared to this, monsoon systems experience much higher discharge ranging between 23.3 and 82.5 m³s⁻¹ for Gangotri glacier (Ramanathan 2010) and 2.4–8.4 m³s⁻¹ for Dokriani glacier (Kumar et al. 2014). In the Ganglass catchment, the high TDS flux despite the low TDS during the peak melt period is attributed to high discharge during this period (Singh et al. 2014; Yde et al. 2014).

Cation weathering rate in the upper Ganglass catchment is 778 meq m⁻² a⁻¹, which is within the range of reported values (660–4,200 meq m⁻² a⁻¹) for the Himalayan glacial basin (Hodson et al. 2002). The significantly higher values of cation weathering rate in glaciers like Dokirani (Table 4) are the result of higher discharge from the monsoonal effect (Hodson et al. 2002). It has been reported that there is more intense weathering in a glaciated catchment compared to a non-glaciated one (Reynolds & Johnson 1972; Collins 1979). The lower glacierized area (Table 4) leads to lower weathering in the present study. The geology of the basin is composed of quartz bearing rocks (quartz, diorite, granodiorite, monzodiorite, monzonite, granite) adding further resistance to weathering.

Not much research has been done to distinguish between glacier and snow cover contribution to the discharge and ion concentration in the runoff of Himalayan basins, leading to misrepresentation of snow melt contribution as glacier melt (Jeelani et al. 2012). In this context, the present study represents the hydrochemical comparison of a snow-dominated basin with a glacier-dominated basin, as shown in Table 5. The biggest difference observed here is the lower ionic concentration of the meltwater from the snow-fed catchment during the major period of the melting season. Second is the higher ionic concentration of snow meltwater reported during the late melt period. In monsoon glacial systems, this occurred either during the early melt period or during the peak melt period (Table 5). A comparison of major cation (Ca) and anion (HCO₃) reveals this characteristic difference. It is also noted that the lowest elemental concentration of the snow-dominated system is lower than the glacier system except for a few exceptions. However, other hydrochemical characteristics of both systems look very similar. Both the systems have a dominance of Ca, HCO₃ and SO₄ as the dominant ion (Singh et al. 1998; Hasnain & Thayyen 1999; Ahmad & Hasnain 2001; Kumar et al. 2009) and equivalent ratios of Ca + Mg/TZ⁺ (0.8–0.9) and Na + K/TZ⁺ (0.1–0.12) are similar to different glaciers (Singh & Hasnain 1998; Kumar et al. 2009). A Piper plot also indicates similar results for snow and glacier dominant systems. The glacier as well as the snow system produces dilute waters but the glacier system has a higher component of subglacial flows, which is chemically enriched due to higher residence time and close contact with high sediment load, resulting in higher ionic concentration of glacier waters.

The present study has been conducted in the Ladakh cryospheric system (cold-arid glacio-hydrologic regime of Himalaya), making the study unique in its kind as very few studies are reported from this system. The entire catchment experiences snowfall in winter and snow melt acts as the major water source in the catchment in summer months. Meltwater is slightly acidic to neutral with Ca and HCO₃ as the major ions throughout the melt period. The high ionic ratios of (Ca + Mg)/TZ⁺ and (Ca + Mg)/(Na + K) with low ratio of (Na + K)/TZ⁺ indicates solute enrichment of snow meltwater by interacting with rock through the process of carbonate weathering followed by silicate weathering. Further dominance of carbonate weathering over silicate weathering in this catchment has been suggested by a correlation matrix (very strong correlation (r > 0.7)) between Ca and HCO₃⁻, factor analysis and Piper diagram. Low NO₃ and Cl concentrations in the meltwater runoff from the cold-arid catchment of Ladakh suggest either a minor contribution from atmospheric sources or its retention in the watershed. Variation in the concentration of most of the dissolved solutes has been noted during early, peak and late melt periods. The discharge–EC relationship clearly shows their inverse relationship for all phases of melt period. The EC along with other dissolved solutes are lowest for the peak melt period suggesting the overwhelming contribution of snowmelt with lower residence time in contact with the rock/sediment. Significantly higher EC during the late melt period indicates higher rock water interaction and probable ground ice melt contribution during this period. The marked variation in EC throughout the melt period highlights the significance of data generation during the whole melting season even in the cold-arid system. The cation denudation rate of the cold-arid regime is found to be 778 meq m⁻² Yr⁻¹ which is significantly lower than the values reported from the monsoon regime, with values as high as 4,160 meq m⁻² Yr⁻¹. Intra-annual variation in the TDS flux (0.64–3.12 ton day⁻¹) is mainly controlled by discharge. The low denudation rate and TDS flux in the upper Ganglass catchment as compared to other glacial basins shows a significant difference between snow–glacial processes. Indications of the contribution of ground ice melt in the cold-arid system runoff that emerged in this study requires in-depth study in the future.

This work has been supported by the Department of Science and Technology (DST) and Council of Scientific and Industrial Research (CSIR). The authors are grateful to SES, Jawaharlal Nehru University for providing the required analytical facility. Thanks also to the Director NIH for his support and encouragement. Assistance by Shailesh for plotting the Piper diagram and help from Parmindra Ola and Naveen Pandey during analysis work is also acknowledged. The authors sincerely thank the two anonymous reviewers for their very constructive and useful comments.