Using the multivariate statistical analysis for understanding of chemical composition of saline lakes in south part of Western Siberia, Russia

. Major ion and trace elements chemistry of fifty-eight water samples of southern part of Western Siberia (Russia) were analysed with a suite of statistical techniques (using R programming language), in an effort to explain the key processes affecting chemical composition of saline lakes. The database included data about well, river, and lake waters and groundwater. According to performed cluster analysis of chemical data, all lakes in the area can be divided in to 5 main groups and their chemical composition very similar to rivers and deep groundwater. Shallow groundwaters are more enriched in Ca 2+ and Mg 2+ . Principal component analysis (PCA) identified four main principal components, with the first principal component (PC1) accounting for 44.3% and representing the process of salinization, and PC2 and PC3 (17.9 and 8.6% of total variance) controlled by water-rock interaction process of carbonates/sulphates precipitation and aluminosilicates formation.


Introduction
Drylands are inhabited by more than 2 000 million people and cover approximately 40 percent of the global land area (excluding Greenland and Antarctica) [1]. Additionally, more than 1400 million people live in drylands of Asia that is 42 percent of the region's population [2]. Nowadays, many saline lakes experience enormous stress not only from natural climatic changes affecting their water balance (many lakes have significantly reduced the water volume, biggest examples are the Aral Sea, Kazakhstan [3] and Lake Urmia, Iran [4], but also from active economic and industrial human intervention in the area of their ecosystems which is accompanied by mismanagement leading to a violation of their geochemical balance.
The catchment basin of saline lakes in Western Siberia in some cases reaches 20 thousand km 2 and often it includes a lot of industrial and technological industries. Being complex dynamic systems that closely interact with the surrounding geochemical environment in the territory of their catchment area, lakes accumulate a huge number of chemical elements, including heavy metals, pesticides, etc., coming from anthropogenic inputs to streams from the entire catchment area [5,6]. Later, these changes in the chemistry of water affect the biological productivity of water bodies, the composition of aquatic biogeocenoses and, consequently, the forage base of fish fauna and aquatic birds [7].
The aim of this study is to examine and interpret the chemical composition of 58 samples of saline lakes, groundwater, cells and rivers located in a steppe area of Western Siberia with total dissolved solids (TDS) varies from 0.5 to 340 g/L and identify the main factors controlling their chemical evolution.

Study sites
The geographical position of Kulunda and Barabinsk steppes is the main factor defining the sharply continental climate of studied region. High instability of climatic factors such as sharp fluctuations in temperature, cold winter with many snowstorms and winds of southwestern direction, hot summers with occasional heavy rains are common features for this region. The average annual temperature ranges from 0 to 0.9 ° in the north, and from 1.6 to 2.1 ° in the south part. Climatic features are frequent droughts, repeated every 3-4 years.
An extremely low (240-320 mm) uneven distribution of rainfall in the annual cycle is a characteristic, while evaporation rates can reach more than 600 mm. Small daily summer precipitation contributes to rapid evaporation without penetrating deep into the soil. Thus, infiltration and replenishment of groundwater reserves in the summer is completely absent. The main flow of atmospheric precipitation infiltration and surface runoff occurs in the spring [8].

Materials and methods
The collected samples of water were filtered into plastic bottles for analysis of major and trace elements. Before sampling all bottles were prepared to prevent contamination. Besides, all unstable parameters were analyzed in the field: pH, Eh, electric conductivity, temperature.
Major element chemistry was measured in the Fundamental Research Laboratory of Hydrogeochemistry of Education and Research Centre "Water" (Tomsk Polytechnic University, Russia) and in Institute of Natural Resources, Ecology and Cryology RAS. Trace element composition of water was also measured in several laboratories (Institute of Solid State Chemistry and Mechanochemistry RAS and Institute of Geochemistry. AP Vinogradova RAS). Obtained from two laboratories chemical composition were compared to each other in order to receive reliable results [8].
Statistical analysis was conducted using R version 3.5.1 [9] with addition of several packages: ggplot2, gridExtra, ggpmisc, corrgram, dplyr, pcaMethods, hydrogeo, ade4, factoextra, ggbiplot. For performed statistical analysis the following elements were selected to calculate PCA: The dataset used in this analysis constitutes a data frame of 58 observations by 25 chemical parameters: pH, Eh, HCO3, CO3, SO4, Cl, Ca, Mg, Na, Li, B, Al, Si, P, V, Cr, Mn, Fe, Cu, As, Br, Sr, Rb, Th, U. Data normality was evaluated using Kolmogorov-Smirnov (K-S) test before the PCA analysis, and non-normally distributed data were logarithmically transformed. The standardization was also applied to ensure that each variable has equal weight in the statistical analyses.

Results and discussion
The summary of chemical composition of all types of water samples is presented in Table  1. According to the Piper diagram (Fig. 1) the chemical composition of groundwater of the studied region is different from the rest of the samples, especially in cation composition: they are more enriched in Ca 2+ and Mg 2+ content ( Table 1). The samples from rivers and deep (confined) groundwater are very close in their chemical percentage composition. According to performed cluster analysis of chemical data, all lakes in the area can be divided in to 5 main groups: clear soda lakes with HCO3+CO3>50%, chloride (or sulphate) lakes with high soda amount (HCO3+CO3 up to 30%) and pure chloride (or sulphate) lakes with Cl or SO4 ion dominance, accordingly.  The results of PCA showed that the first four factors account for 77.3% of the total variance in studied dataset. The first two factors, respectively, explain 44.3% and 17.9% of the variance, and therefore explain most of the variance in the dataset, while factors 3 and 4 explain 8.6% and 6.5%. As can be seen from Fig. 2(a) the PC1 mostly explained by evaporation processes and correlated mostly with elements which accumulate in the solution by increase of salinity (Na, Cl, SO4, Br, Li, B, etc.). Usually these elements do not form their own mineral assemblages and migrate in pure ion form. Fig. 3(a, b) clearly shows how chemical types are replace each other from HCO3-Ca(Na) to Cl-Na and (Cl)-SO4-Na with 3 outliners of pure soda lakes (Petuhovo, Tanatar-4 and Tanatar-6). However, as can also be seen from Fig.2 (a, b), evaporation factor is not the only one that affects the chemistry of studied waters. An important factor in the formation of hydrochemistry of lakes is the mineral precipitation associated with the saturation of water by different mineral phases. The second and the third principal components are closely connected with the processes of water-rock interaction. PC2 is highly correlated with the amount of HCO3/CO3 and Ca/Mg/Sr that can be explained by the specifics of their precipitation process. According to thermodynamic calculations (Fig. 4), carbonate sedimentation is ubiquitous for the majority of studied waters. Therewith, the gypsum mineral formation stage is not typical for them mainly due to the complete removing of calcium by carbonate minerals. However, since the sulphate content in the solution is still limited by celestite and thernadite precipitation, water equilibrium with mirabilite is not achieved. The third PC3 is explained by aluminosilicate formation and highly correlated with SiO2 content.

Conclusions
The obtained results provide the complex version of chemical composition of groundwater and saline lakes required for monitoring and forecasting of the ecological status of natural objects in the south of Western Siberia. All the lakes in the area are divided in to 5 main groups: clear soda lakes with HCO3+CO3>50%, chloride or sulphate lakes with high soda amount (HCO3+CO3 up to 30%) and pure chloride or sulphate lakes with Cl or SO4 ion dominance, accordingly.
Principal component analysis identified three main principal components, with the first principal component (PC1) representing the process of salinization, and other are controlled by water-rock interaction process of carbonates/sulphates precipitation (PC2) and aluminosilicates (PC3) formation.
The data assessment and study were founded by Russian Science Foundation (RSF, Project 17-77-10086).