Hydrochemistry of the Lhasa River, Tibetan Plateau: Spatiotemporal Variations of Major Ions Compositions and Controlling Factors Using Multivariate Statistical Approaches

: Spatiotemporal variations of the hydrochemical major ions compositions and their controlling factors are essential features of a river basin. However, similar studies in the southern Tibetan Plateau are relatively limited. This study focuses on the chemical compositions of the dissolved loads in the Lhasa River (LR) in the southern Tibetan Plateau. Two sampling campaigns were conducted during the rainy and dry seasons across the LR basin to systematically investigate the spatiotemporal variations of water chemistry and sources of the dissolved loads. The results show that the river water possesses slight alkalinity with an average pH of 8.05 ± 0.04. Total dissolved solids (TDS) and oxidation-reduction potential (ORP) range widely from 39.8 mg/L to 582.6 mg/L with an average value of 165.6 ± 7.7 mg/L and from − 9.4 mV to 295 mV with a mean value of 153.7 ± 6.9 mV, respectively. The major cations follow the decreasing order of Ca 2+ , Mg 2+ , Na + , and K + while HCO 3 − , SO 42 − , Cl − , and NO 3 − for anions. Ca 2+ and Mg 2+ account for 87.8% of the total cations, while HCO 3 − and SO 42 − accounts for 93.9% of the total anions. All the major ions show higher concentrations in the dry season. NO 3 − , HCO 3 − , and Mg 2+ show signiﬁcant spatial variations due to the inﬂuence of basin lithology and anthropogenic activity. Multi-variables statistical analysis reveals that the mechanisms controlling the LR hydrochemistry are mainly carbonate weathering followed by silicate weathering. Geothermal springs and anthropogenic activities also play crucial roles in altering river water ions composition in the middle stream and downstream. The relatively high NO 3 − value (3 ± 0.2 mg/L) suggests water quality will be under the threat of pollution with the increase of anthropogenic activities.


Introduction
Rivers play an important role in industries and domestic water usage as well as irrigation [1]. Thus, the assessment of the water quality is a prerequisite for effective and reasonable water management. The river chemical compositions reflect the climate and environmental conditions where the river drains [2]. Previous studies suggested that the controlling factors of surface water chemistry are extremely complex and show spatiotemporal heterogeneity. Controlling factors include many natural processes, i.e., precipitation, parent rock weathering [3][4][5], evapo-crystallization process [6], tectonic movement, geographic and geomorphic conditions [3], and anthropogenic inputs (i.e., mining activities) [1,3], agricultural fertilizer [7,8], and city sewage [9]. Numerous studies

Study Area
The LR originated from the Nyenchen Tanglha Mountains, which is the largest tributary of the Yarlung Tsangpo ( Figure 1). The LR basin is located at the southern Tibetan Plateau and extends from 90.1 • E to 93.3 • E and 29.3 • N to 31.3 • N with an area of 3.26 × 10 4 km 2 . Elevation ranges from 3575 m to 7085 m with a mean altitude of 4500 m. Controlled by the South Asian monsoon, the LR basin is characterized by a temperature semiarid monsoon climate with perennial annual precipitation of 440 mm [26]. Precipitation displays huge spatiotemporal heterogeneity. The whole year can be divided into two seasons: dry season occurring from October to May of the following year; rainy season initiating from June to September when almost 90% of rainfall drop during this period [39,40]. The precipitation gradually decreases from northeast to southwest. The annual average air temperature is −1.7 to 9.7 • C [40].
Vegetation types display significant spatial differences with alpine meadows and subalpine steppes distributing in the up-and middle-stream whereas shrubs and forest in the lower reaches of the LR [38]. Soil types in the area are mainly alpine meadow soil, alpine desert soil, and sub-alpine steppe soil [40]. Geographic features are characterized by high mountains and valley plain [37]. Quaternary strata are widely distributed and composed of alluvial and diluvial deposits [36]. The lithology of the LR basin is dominated by the Mesozoic limestone and metamorphic sandstone and Himalayan granite [36] ( Figure 1B).
The LR is fed by precipitation, glacier-snow meltwater, and groundwater [32]. In the upstream of the LR, the main anthropogenic activity is pasture husbandry [38]. The croplands and cities are mainly situated in the river valley of the lower reaches. There are abundant geothermal springs distributed within the LR basin along the Nyenchen Tanglha Mountains foothill [28,29] (Figure 1). The sampling sites (marked by solid circles) during the rainy and dry seasons. The red short lines are nodes that divide the LR channel into the upstream, middle stream, and downstream, respectively. The red ellipse means the source area of the LR. The red numbers mean the tributaries of the Lhasa River, which are: 1 Maiqu; 2 Sangqu; 3 Wululongqu; 4 Xuerong Tsangpo; 5 Mozhuqu; 6 Pengboqu; 7 Duilongqu. The red dashed line is the geothermal water distribution zone. (B) Lithological map of the Lhasa River basin. The lithology data was obtained from the global lithological map (GLiM).

Sampling and Analyzing
Water samples of the LR basin were collected in August and November 2019, which belong to the rainy season and dry season, respectively. A total of 160 samples were collected at 80 sites during two seasons. Surface water samples include glacier meltwater, snow, tributaries, and mainstream. The basic hydrochemistry parameters were measured in situ with the portable multi-parameter meter (HANNA, Co., Laval, QC, Canada), including water temperature (Tw, ±0.15 • C), electrical conductivity (EC, ±1 µS/cm), dissolved oxygen (DO, ±1.5% to ±3%), pH (±0.02) and oxidation-reduction potentiometer (ORP, ±1.0 mV). All devices were calibrated with standard solutions before measurement. Samples were filtered with 0.45 µm Nuclepore filters (Advantec, Co., Tokyo, Japan) and injected into 50 mL polyethylene bottles for major anions (i.e., chloride (Cl − ), sulfate (SO 4 2− ) and nitrate (NO 3 − )) and cations (i.e., calcium (Ca 2+ ), magnesium (Mg 2+ ), sodium (Na + ), and potassium (K + )) analysis. The snow and glacier ice samples were collected into plastic bags by wearing polyethylene gloves and then melted at room temperature and finally transferred into 50 mL polyethylene bottles. All of the sample bottles were sealed with parafilm and kept in a frozen environment until follow-up analysis.
Major ions (exclude HCO 3 − ) were measured with ion chromatography (ICS-1100 Dionex. Co., Sunnyvale, CA, USA) and soluble silicate was determined by flow injection analysis catalytic spectrophotometric method (QuickChem ® Method 31-114-27-1-D, Loveland, CO, USA) in School of Environmental Science and Engineering, Southern University of Science and Technology, China. The detection limits are 100 µg/L and 50 µg/L for major anions and cations, respectively. HCO 3 − was titrated with 0.16 N H 2 SO 4 2− cartridge (Hach Co., Loveland, CO, USA) in situ by the following method. First, 100 mL water sample was poured into a clean, 250-mL Erlenmeyer flask. Second, one phenolphthalein indicator powder pillow was poured into the sample and swirled to mix. Third, the sample was titrated with a sulfuric acid solution until the color changed from pink to colorless (pH ≈ 8.3). This value indicates the phenolphthalein alkalinity in the sample. Fourth, a bromcresol green-methyl red indicator was added and the titration continues until the color changed from green to red (pH ≈ 4.3 to 4.9). This value indicates the total alkalinity. Finally, hydroxide, carbonate, and bicarbonate alkalinity were calculated based on the alkalinity relationships. All of the samples possess the charge imbalance of less than 10% and most of them possess the charge imbalance of less than 5%.

Statistical Analysis
t-test and one-way analysis of variance (ANOVA) were conducted to compare the difference of major ions between rainy season and dry season, and among upstream, middle stream, and lower stream of the LR, respectively [41,42]. Multivariate statistical analysis is an effective method in analyzing the relationships of hydrochemistry parameters and has been widely used [43][44][45][46]. The Spearman correlation analysis was used to identify the relationships among hydrochemical variables [8,43]. Principle component analysis (PCA), a typical dimensionality reduction method used to extract the explanatory power of numerous variables [46][47][48][49], has been widely used to distinguish the origin of major ions [46,47,50]. The above multivariate statistical analysis was performed by the R program.

General Hydrochemistry
Concentrations of major ions and basic water parameters of the LR and adjacent rivers are shown in Table 1. The mean value of pH is 8.1 indicating a slightly alkaline water environment. The T w of LR varies from 0.1 to 19.7 • C during the sampling periods. TDS (TDS = ∑(Ca 2+ + Mg 2+ + K + + Na + + HCO 3 − + SO 4 2− + Cl − + NO 3 − + SiO 2 , in mg/L) ranges widely from 39.8 to 582.6 mg/L with a mean value of 165.6 ± 7.7 mg/L. The LR displays a relatively smaller TDS compared with surrounding rivers and a greater value than the global mean value of 120 mg/L [51] and the Yarlung Tsangpo 112 mg/L [13]. ORP reflects an oxidized environment with a mean value of 154 ± 7 mV. The smallest value of ORP (−9 mV) was shown at the lower reach suggesting the reducibility of tributary water.
Calcium is the most abundant cation ranging from 6.3 to 111.1 mg/L with a mean value of 27.5 ± 1.3 mg/L, followed by magnesium (5.0 ± 0.4 mg/L) and sodium (5.2 ± 0.6 mg/L). Potassium possesses the least proportion with merely mean content of 1.2 ± 0.1 mg/L. The relative proportions of cations can be presented in a ternary diagram ( Figure 2). Almost all the points gathered towards the Ca 2+ apex, which further illuminates calcium is the dominant cation accounting for 23 -88.1% (68.5%) on an equivalent basis. Mg 2+ and Na + + K + account for 5.1 -48.9% (19.3%) and 2.6 -66.7% (12.2%) of total cations, respectively. Note. B.D. means below detection limit; a means concentration of HCO 3 − + CO 3 2− ; units for major ions concentration are in mg/L and EC in µs/cm. TDS = ∑ (Ca 2+ + Mg 2+ + K + + Na + + HCO 3 − + SO 4 2− + Cl − + NO 3 − + SiO 2 ). Bicarbonate and sulfate are the dominant anions with mean concentrations of 68.3 ± 3.3 mg/L and 45.4 ± 4.0 mg/L, respectively. Chloride displays a mean value of 3.1 ± 0.5 mg/L with the largest value (46.8 mg/L) shown in the downstream, which is affected by the surrounding geothermal springs [28]. The relatively high nitrate concentrations (3.0 ± 0.2 mg/L) are almost three times the global average value (1 mg/L, [51,52]). In the anion ternary diagram, all the sampling points distribute along the HCO 3 − axis. SO 4 2− and HCO 3 − together account for 72.3 -99.4% (93.9% on average) of total anions on an equivalent basis, implying the major anions are SO 4 2− and HCO 3 − . The water chemistry data of the LR is comparable to the other rivers derived from the Tibetan Plateau, indicating the LR water is relatively pristine. The LR possesses a relatively lower TDS compared with other rivers originated from the Tibetan Plateau (Table 1) whereas greater TDS than the global river mean value (120 mg/L, [51,52]). What is noteworthy is that NO 3 − of the LR is the highest among the surrounding rivers as well as the global mean value of 1 mg/L [51,52].
Surface waters in the LR basin show low mineralization (Tables 1 and 2). The mean values of all the chemical parameters are within the range in values of water chemistry data in the Chinese State Standard (CSS) [53] and World Health Organization (WHO) [54] for drinking water (Table 1). There is one sample from downstream with SO 4 2− exceeding the maximum limit, which should be caused by geothermal springs inputs.

Seasonal Variations of Hydrochemistry
In the rainy season, the mean value of surface water temperature is 13.7 ± 0.3 • C while it is only 5.2 ± 0.5 • C during the dry season. The ORP values reflect oxidized environments for water in both sampling seasons, with a mean value of 92 ± 4 mV in the rainy season and increased up to 215 ± 8 mV in the dry season. TDS displays considerable variations from 39.8 to 582.6 mg/L with a mean value of 138.7 ± 9.0 mg/L for the rainy season and from 47.3 to 557.4 mg/L with an average value of 192.5 ± 11.8 mg/L for the dry season, respectively. These findings are in accordance with the results around the Himalaya Mountains regions [8,55]. The seasonal variation of TDS results from precipitation and meltwater dilution [8]. Ca 2+ is the most abundant cation in both sampling periods and the mean value of the dry season is 1.45 times higher than that of the rainy season ( Figure 3). K + shows the lowest value of 0.9 ± 0.1 mg/L and 1.4 ± 0.1 mg/L for the rainy and dry seasons, respectively. Concentrations of cations follow the decreasing order of Ca 2+ > Mg 2+ >Na + > K + during both sampling periods. Concentrations of anions follow the decreasing order of HCO 3 − > SO 4 2− > NO 3 − > Cl − in the rainy season while change to HCO 3 − > SO 4 2− > Cl − > NO 3 − in the dry season. It is worth noting that NO 3 − is an important hydrochemistry parameter that illustrates the interference of anthropogenic activities on streams. The concentrations of NO 3 − range from 0.02 to 10.7 mg/L and 0.02 to 9.7 mg/L with mean values of 2.7 ± 1.7 mg/L and 3.4 ± 2.3 mg/L for the rainy season and the dry season, respectively, and display significant temporal variations.
The T-test results show that all the major cations and anions show a significant temporal difference with p < 0.05 (Figure 3). It can be concluded that concentrations of major ions in the dry season are generally much higher than those of the rainy season, which may be caused by the integrated dilution effects of heavy precipitation and large amounts of glacier meltwater [8].

Spatial Distribution of Major Ions
Across the LR basin, the mean pH was found to be alkaline with the highest pH (9.4) noted from the upstream and the lowest pH (5.5) found in the Mozhu Qu (Table 2). Water samples from the middle stream possess higher TDS than those of upstream and downstream (Table 1). However, the highest (582.6 mg/L) and lowest TDS (39.8 mg/L) values are from the downstream. Several samples collected adjacent to the Nyenchen Tanglha Mountains show significantly lower TDS, which are affected by the dilution effect of the glacier meltwater.
One-way ANOVA analysis indicates that HCO 3 − , Mg 2+ , and NO 3 − show significant spatial variations with p < 0.05 among different reaches while other ions show no significant spatial variations with p > 0.05 (Figure 4). The concentration of K + is the lowest among the cations and its distribution is relatively even across the LR basin except for the largest value found in Wululong Qu (8.4 mg/L). Na + and K + display extremely high concentrations upstream of Sangqu, Wululongqu, and Duilongqu (Table 2), which is belonging to the geothermal springs belt [3]. Ca 2+ (Table 1). It should be noted that the mean value of NO 3 − is especially high in sites around cities and towns, which can be attributed to agricultural fertilizer usage and domestic wastewater discharge [38]. The elevated NO 3 − concentrations will threaten drinking water safety.

Mechanisms Controlling the Major-Ion Chemistry
The major sources of dissolved loads can be classified into atmospheric input, parent rock weathering, and anthropogenic inputs [6]. The surface water samples collected from both seasons are characterized by the medium TDS (165.6 ± 7.7 mg/L) and low ratios of Na + /(Na + + Ca 2+ ) and Cl − /(Cl − + HCO 3 − ) (<0.5), which demonstrate the controlling role of rock weathering (Table 3). Chemical weathering of different lithologies yield complicated dissolved ions. For instance, carbonates (calcite and dolomite) weathering produce Ca 2+ , Mg 2+ , and HCO 3 − ; silicates (albite, K-feldspar, Ca-plagioclase, and Olivine) weathering generates Na + , K + , Ca 2+ , Mg 2+ , Si, and HCO 3 − , and evaporates dissolution (halite and gypsum) are mainly the sources of Ca 2+ , Mg 2+ , Na + , K + , Cl − , and SO 4 2− [10]. Thus, mixing diagrams of Na-normalized ions molar ratios (meq) have been used to display various endmember mixing [8,42,51]. Most of the samples collected from both seasons cluster towards the carbonates end-member, which implies the dominant role of carbonate (calcite and dolomite) weathering ( Figure 5). Meanwhile, there are still some samples collected from the middle stream and downstream during the dry season gathering towards silicates and geothermal springs end-members, suggesting the origins of ions from silicate weathering and geothermal springs discharge. Note. Values expressed as mean ± Se, all ratios derived from meq concentrations and a ratios from mmol/L concentrations. Total dissolved cations: TZ + = Na + + K + + 2 × Mg 2+ + 2 × Ca 2+ and total dissolved anions: Ionic ratios can also further illustrate the origin of major ions yielded by chemical weathering (Table 3). Generally, carbonates are susceptible to weathering and can alter easily the hydrochemistry even though few carbonate rocks are presented in the bedrock [3,56]. The high equivalent ratios of (Ca 2+ + Mg 2+ )/(Na + + K + ) (7.8 ± 0.7 to 13 ± 1.6) and (Ca 2+ + Mg 2+ )/TZ + (0.8 to 0.9 ± 0.01) also illustrate the dominant role of carbonates (calcite and dolomite) weathering on the LR water chemistry.
The equal ratios of Na + and Cl − reflect Na + originates from the halite dissolution [57]. However, the concentration of Na + (in meq/L) is much higher than the concentration of Cl − ( Figure 6A and Table 3). The scatter plot of Cl − and Na + exhibit that most samples lie above the 1:1 line, which implies the additional inputs, such as weathering of silicates and ion exchange [37] as well as geothermal springs input [3,28]. The excess Na + can be balanced by Si ( Figure 6B). It can be seen that most of the samples lie on the 1:1 line, suggesting that Na + mainly originates from silicate weathering. There are still several points that lie above the 1:1 line, which may imply the additional Na + originates from geothermal springs input. End-members of evaporates (Evap), silicates (Sil), and carbonates (Carb) are attained from Gaillardet et al. [51]. End-member of geothermal water (GeoW) was obtained from Guo et al. [28,29]. End-member of geothermal water (GeoW) was obtained from Guo et al. [28,29].
is an important indicator of proton source required for carbonate weathering [8]. If the C-ratio is less than 0.5, the coupled reaction of the sulfide oxidation and carbonate weathering are illuminated. When the C-ratio is close to 1, the carbonation reaction of atmospheric CO 2 is the dominant proton source [8]. The equivalent ratio of HCO 3 − /(HCO 3 − + SO 4 2− ) is >0.5 for the whole LR basin during the rainy season and for the middle stream and downstream during the dry season, which indicates that the proton is originated from atmospheric CO 2 . Whereas the proton of the upstream during the dry season is sulfide oxidation for the low equivalent ratio of HCO 3 − /(HCO 3 − + SO 4 2− ) (0.48 ± 0.03). According to the lithological map ( Figure 1B), the lithology of the sampling sites in upstream is mixed sedimentary rocks. The equivalent ratio of Ca 2+/ SO 4 2− is greater than 1 ( Figure 6C) further confirming that CO 2 dissolution is the major proton source.
The equivalent ratios of (HCO 3 − + SO 4 2− )/(Ca 2+ + Mg 2+ + Na + ) from different reaches are close to 1, which implies the dominant role of carbonate weathering and the limited role of silicate weathering on the LR hydrochemistry ( Figure 6D). The hydrochemical characteristics of groundwater in the LR are characterized by moderate TDS and the major ions originated from the weathered carbonates and silicate [36]. There are intensive hydraulic connections between groundwater and LR water and they are mutually transformed [58].
Overall, the mechanisms controlling the major ions chemistry display spatial variations. In the upstream, the major ions composition is dominated by carbonate weathering. While in the middle stream and downstream, besides lithology, anthropogenic activities and geothermal springs water input also play important roles in major ions chemistry.
The major ions of the LR and the rivers draining the Himalaya Mountains are both originated from chemical weathering. Carbonate weathering plays the dominant role in river hydrochemistry [56,59,60]. Previous studies have demonstrated that the rivers draining the Himalaya Mountains exhibit higher chemical weathering rates than the rivers originated from the Tibetan Plateau [3,55]. This can be attributed to heavy monsoon precipitation and rapid tectonic uplift along the Himalaya Mountains whereas low precipitation and cold environment on the Tibetan Plateau [3,10].

Multivariate Statistical Methods
The correlation matrix of water chemistry indices was used to investigate the dependency of one hydrochemistry parameter on the other. The correlations of geographical factors, water quality indices, and major ions concentrations are presented in Figure 7. The results indicate strong positive correlations between EC and TDS, Ca 2+ , Mg 2+ , HCO 3 − , and SO 4 2− . The positive correlations among Ca 2+ , Mg 2+ , and HCO 3 − in both sampling seasons suggest the contribution of the carbonates weathering to these ions. Ca 2+ , Mg 2+ , and SO 4 2− show positive correlations that illustrate probably the dissolution of gypsum or carbonate dissolution caused by sulfide oxidation [61]. However, Cl − is not significantly correlated with SO 4 2− in the rainy and dry seasons (Figure 7), which implies that the dominant origins of SO 4 2− is from pyrite oxidation [3]. The source of SO 4 2− for the rivers draining the Himalaya Mountain is also originated from sulfide oxidation [3,10].
The significant positive correlations among Cl − , Na + , K + , and SiO 2 during both sampling seasons imply the same source of these ions, possibly from the effect of weathering of silicate and geothermal springs inputs. Furthermore, elevation is strongly negatively associated with NO 3 − since anthropogenic activities are mainly carried out in the low elevation regions, which further suggests the effect of anthropogenic activities on the ion composition of water.
PCA was used to further reduce the dimensionality of the hydrochemical dataset containing considerable interrelated variables while retaining the variation shown in the dataset [46]. Four principal components (PCs) with eigenvalue > 1 were extracted explaining 70% of the total variance (Table 4 and Figure 8). PC1 has strong loadings on HCO 3 − , SO 4 2− , Ca 2+ , Mg 2+ , TDS, and EC accounting for 34.88-36.44% of the total variance, which indicates carbonate weathering is the dominant origin of these ions. PC2 explains 18.33-24.49% of the total variance with strong loadings of Na + , Cl − , SiO 2 , and K + , which suggest the mixing contribution of silicate weathering and geothermal springs [3]. PC3 is the factor of topography and climate with strong loadings of Elevation (Ele) and T w , which explain 10.83-13.09% of the total variance. PC4 explains only 8.25-8.89% of the variance with strong loadings of NO 3 − , which can be attributed to the effect of anthropogenic activities. In general, the dominant source of ions is the parent rock weathering, followed by the geothermal springs' inputs and monsoon climate, as well as anthropogenic activity.

Conclusions
This study investigates the hydrochemical characteristics of the LR during the dry and rainy seasons. The LR water quality is relatively pristine and possesses a slightly alkaline aquatic environment with an average pH of 8.1 and a mean TDS of 165.6 ± 7.7 mg/L during the sampling period. Ca 2+ and Mg 2+ are the dominant cations which account for 80% of the total cations. HCO 3 − and SO 4 2− are the dominant anions contributing 90% of the total anions. All of the major ions show significant seasonal variations with lower concentrations in the rainy season owing to the coupled dilution of the intense precipitation and glacier meltwater. The temporal variation of NO 3 − and SO 4 2− is significant in both seasons. HCO 3 − , Mg 2+ , and NO 3 − show significant spatial variations (p < 0.05) across the LR basin owing to the complex geological settings and anthropogenic activities.
The ternary plot, mixing diagram, and multivariate statistical analysis illustrate lithology and monsoon climate jointly drive the temporal-spatial variance of the hydrochemistry parameters. The dominant mechanism affecting the LR hydrochemistry is the parent rock weathering followed by the topography and monsoon climate as well as anthropogenic activities. NO 3 − is significantly negatively associated with elevation, which is caused by anthropogenic activities carried out in the low elevation regions. SO 4 2− is positively correlated with Ca 2+ , Mg 2+ , TDS, and EC. Geothermal springs are also an important endmember enriched in Cl − , SO 4 2− , Na + , and Ca 2+ . The mechanisms controlling the major ions chemistry show spatial variations. Carbonate weathering is the major source of hydrochemistry in the upstream, middle stream, and downstream of the LR, respectively, followed by silicate weathering. Geothermal springs input and anthropogenic activities are also important end-members for major ions in the middle stream and downstream. There is no doubt that hydrochemistry will be under threat with the increase of anthropogenic activities, especially in the middle reaches and downstream of the LR.